International Journal of Intelligent Systems and Applications in Engineering

Ontology-based Multi-Agent System on Fuzzy Markup Language in Healthy Lifestyle

Jayaprakash Sunkavalli — 2024-03-26

The best ways to avoid illness are to lead a healthy lifestyle and eat a balanced diet. A healthy lifestyle is centered on good eating practices. A person's risk of illness will rise if they consistently consume too little or too much. Thus, the development of balanced and healthful eating habits is crucial to the prevention of disease. To record and depict the agents as well as their actions, which give them the capacity for reasoning, we also propose an ontology-based category knowledge and context framework. The procedure has been helped to accomplish that goal by the introduction of numerous strategies and technology. One technique that is gaining popularity to support knowledge exchange within organizations is ontology, which is a method of representing knowledge. This work offers an ontology-based multi-agent system (OMAS) for diet health evaluation that consists of a fuzzy inference agent, a semantic generation agent, and an individual information agent. The users are then asked to enter the foods they have consumed. Lastly, subject matter experts construct the ontologies for food and personal profiles. The OMAS's knowledge base and rule base are described using fuzzy markup language (FML). The primary output of basic research in healthcare informatics is the development of domain ontologies and problem-solving techniques. Consequently, our scientific community has to give these ideas more consideration.

Real-Time Imbalance Liver Tumor Sensor Databases: A Deep Classification Framework with Ensemble Feature Extraction, Ranking, and Probabilistic Segmentation for Efficient Analysis

N. Nanda Prakash — 2024-03-26

As the size of liver tumor image databases increases, it becomes challenging to enhance the true positive rate of traditional prediction approaches due to the high majority-minority ratio and noise. However, while 3D convolutions have the potential to fully leverage spatial information, they also come with high computational costs and require significant GPU memory usage. On the other hand, 2D convolutions are limited in their ability to utilize the information contained in the third dimension. Missing feature values, feature noise, and Imbalanced liver classes are some of the significant factors that can impact the quality of input data. The quality of imbalance data significantly impacts the efficiency of classification approaches, making it necessary to ensure high-quality input data to achieve optimal results. Therefore, to ensure high-quality predictions on imbalanced liver datasets, models need to be optimized. Sensors are commonly used to collect and measure physical parameters, and they can be used to obtain liver tumor data for the proposed model. In this work, medical imaging sensors such as CT (computed tomography) machines are used to capture detailed images of the liver and identify potential tumors. Therefore, sensors play a crucial role in the proposed model by providing the necessary data to extract features, segment the liver and detect tumors accurately. In this work, an optimized k-joint probabilistic segmentation-based ensemble classification model is proposed to address the issues of homogenous and heterogenous liver tumor detection. Additionally, novel image filtering, feature extraction and ranking approaches are proposed to improve the imbalanced liver tumor regions for classification process. The experimental results demonstrate that the proposed classification model based on k-joint probability segmentation has significantly improved the accuracy, recall, precision, and AUC compared to the existing models.

Improvised Swarm Based Discrete Data Mining Approach for High Utility Item Sets

Raja Rao Budaraju — 2024-03-26

Data mining techniques uncover valuable patterns hidden inside extensive databases to assist decision support systems in different practical applications. Association rule mining analyzes the transaction database to recognize patterns and provides insights into client behavior. Frequent itemset mining (FIM) detects a group of items that are commonly purchased together. A significant limitation of FIM is its disregard for the item's significance. The significance of an item is crucial in a practical application. Hence, it is imperative to identify the critical set of items that yields substantial profits, referred to as the HUIM (High-Utility Itemset Mining) problem. Various techniques may be employed to identify high utility itemsets from a transaction database. The HUIM approaches that employ the Utility list are relatively new and exhibit superior performance in terms of memory consumption and execution time. Primary constraint of these algorithms is the execution of expensive utility list join operations. This work presents a highly efficient optimization approach based on swarm intelligence for addressing the HUIM problem. Furthermore, the suggested method's execution time is assessed. Additionally, it is compared to relevant and advanced current approaches. Extensive tests were carried out on accessible benchmark datasets demonstrate that the suggested swarm-based methodology outperforms state-of-the-art approaches.

Machine Learning with IoT Enhancing Car Performance through Supervised Algorithms for Vehicle Automation

Elangovan G. — 2024-03-26

The present research explores the combined application of supervised learning algorithms and the Internet of Things (IoT) to improve automotive performance in the context of vehicle automation. Our study makes use of neural networks, decision trees, and support vector machines along with a variety of datasets, well-placed sensors, and communication protocols. Across ten trials, the selected algorithms consistently displayed excellent performance, generating accuracy values ranging from 91.7% to 93.5%, precision values between 93.7% and 94.8%, recall values spanning from 89.8% to 91.7%, and F1 scores ranging between 91.5% and 93.4%. These striking results underline the potential of this integrated strategy to transform driving experiences, increase safety, and contribute to the continued growth of intelligent vehicle systems. This research not only lays the framework for new developments in the automotive sector but also demonstrates the revolutionary impact of advanced technology on the landscape of modern transportation.

Privacy Preserving Cyber Security System Framework for Secure Cloud-based Medical Data Transactions

Sunil D. Kale — 2024-03-26

With the increasing adoption of cloud storage in healthcare, ensuring the privacy and security of medical data has become a critical concern. This paper presents a comprehensive framework for privacy-preserving cybersecurity to secure medical data transactions in cloud storage environments. The proposed framework focuses on preserving the confidentiality and integrity of medical data while facilitating secure transactions. It integrates encryption methods to safeguard sensitive data both at rest and during transmission, guaranteeing that only authorized individuals have the ability to access and decrypt the information. Access controls are implemented to enforce fine-grained permissions, restricting data access to authorized personnel based on roles and privileges. Additionally, the framework includes mechanisms for secure authentication and identity verification to prevent unauthorized access and mitigate the risk of data breaches. Compliance with regulatory requirements such as HIPAA and GDPR is also addressed to ensure that the framework meets the necessary standards for protecting patient privacy. By integrating privacy preservation principles with robust cybersecurity measures, the proposed framework provides a comprehensive solution for securely managing medical data in cloud storage, enhancing trust and confidence in digital healthcare systems.

An Automatic Multi-Variate Multi-Class Feature Extraction, Ranking Based Joint Probabilistic Segmentation and Classification Framework for Multi-Class Liver Tumor Detection

N. Nanda Prakash — 2024-03-26

Automatic multivariate and multi-class tumor detection plays a crucial role in the diagnosis and treatment of large heterogeneous liver databases. However, existing liver segmentation models often encounter challenges such as multi-modal tumor detection, detecting tumors with varying shapes, liver over-segmentation, and difficulty in identifying tumors with different orientations and shapes. Moreover, the presence of noise in excessively segmented border regions can further complicate the segmentation and classification process, leading to inconsistent and inaccurate results. In this work, we propose novel approaches for multivariate liver filtering, multivariate feature extraction, and ranking. We employ efficient multivariate segmentation-based classification methods to enhance the overall detection of multi-modal tumors on large databases. The proposed Multi-variate Liver and Tumor Segmentation and Classification (MCMVLTSC) model efficiently classifies key tumor segmented regions with a high true positive rate and runtime efficiency (ms). To evaluate the performance of our proposed MCMVLTSC model compared to existing models, we utilize various statistical measures on diverse liver imaging databases. Experimental results demonstrate that the proposed model outperforms conventional models in terms of different statistical classification metrics and runtime efficiency.

A Hybrid Approach Using AES-RSA Encryption for Cloud Data Security

Juvi Bharti — 2024-03-26

This paper provides an in-depth analysis of cloud vulnerabilities and explores the role of encryption techniques in addressing security challenges within cloud services. We compare a range of symmetric and asymmetric encryption algorithms and discuss recent advancements in modified encryption models. The study highlights the importance of understanding the encryption landscape, selecting appropriate techniques based on specific requirements and threat models, and keeping up to date with the latest developments in cryptography to ensure data security in the cloud.

Algorithmic Insights into Predicting Hypertension Using Health Data in Cloud-Based Environments

S. V. N. Sreenivasu — 2024-03-26

This exploration examines the use of cutting-edge calculations for anticipating hypertension inside cloud-based health conditions. Utilizing assorted health information sources, including electronic health records and wearables, we investigated the prescient abilities of four key calculations: Strategic Relapse, Random Forest, Backing Vector Machine (SVM), and Neural Network (Multi-facet Perceptron). Our exploratory arrangement included thorough information preprocessing, highlight extraction, and model preparation on an extensive dataset. The Neural Network arose as the best calculation, accomplishing an exactness of 90%, accuracy of 92%, review of 88%, F1 score of 90%, and an AUC-ROC of 0.94. Random Forest and SVM likewise exhibited hearty execution with a precision of 88% and 87%, individually. Calculated Relapse, however less difficult, displayed cutthroat dependability with a precision of 85%. Correlations with related work highlighted the adaptability of the calculations, reaching out past unambiguous medical services spaces. This exploration adds to the more extensive talk on prescient medical services examination, stressing the reconciliation of cutting-edge calculations in cloud-based conditions. Our findings set the stage for subsequent research, which may include the continuous observation of IoT devices and the improvement of profound learning designs, all while recognizing specific constraints like the representativeness of the dataset and the model's interpretability.

Historical Data Mining and Cultural Heritage Inheritance Path Modeling of Traditional Architecture in the Guangfu Region

Jin Ling — 2024-03-26

Historical data mining plays a crucial role in cultural heritage inheritance by uncovering insights from vast repositories of historical information. Through advanced data analysis techniques, such as machine learning algorithms and pattern recognition, historical data mining enables the extraction of valuable knowledge from historical documents, artifacts, and archaeological findings. in historical data mining for cultural heritage inheritance include data quality and accessibility, cultural sensitivity, and interpretation challenges. Historical datasets may suffer from inconsistencies, incompleteness, or biases, posing challenges to the accuracy and reliability of mining results. Additionally, accessing historical data, especially from remote or protected cultural sites, can be challenging due to legal, logistical, or ethical considerations. Cultural sensitivity is crucial, as historical data may contain sensitive or contentious information that requires careful handling and interpretation to avoid misrepresentation or offense. This study explores the application of historical data mining techniques in the context of cultural heritage inheritance, focusing on traditional architecture in the Guangfu region. Leveraging Software-Defined Hierarchical Clustering Path Modeling (SDHCPM), the research aims to uncover underlying patterns and pathways in the evolution of traditional architecture, shedding light on its historical significance and cultural heritage preservation. By analyzing historical datasets related to architectural styles, construction techniques, and socio-cultural influences, SDHCPM facilitates the construction of a path model that traces the development of traditional architecture over time. Through this approach, the study seeks to enhance our understanding of the cultural heritage of the Guangfu region and provide valuable insights for heritage conservation and revitalization efforts. the average clustering coefficient for traditional architecture in the region is found to be 0.75, indicating a high level of architectural coherence and cultural continuity.

Rural Landscape Pattern Analysis and Optimization Model Construction Based on Remote Sensing Technology

Shuai Xiao — 2024-03-26

Rural landscape pattern analysis involves the examination of the spatial arrangement and composition of land cover types in rural areas. in rural landscape pattern analysis encompass challenges such as data availability, scale discrepancies, and methodological complexities. Limited access to high-resolution spatial data, particularly in remote or developing regions, can impede accurate analysis and interpretation. Scale discrepancies between the spatial extent of data sources and the ecological processes being studied can also affect the reliability of findings. Hence, this paper proposes Genetic Optimized Stimulated Annealing Multi-Spectral (GSA-MS) for the pattern analysis. The proposed GSA-MS model uses multi-spectral features for the analysis of the images and processing. With the GSA-MS model features are extracted in the rural images for the estimation of patterns. With the GSA-MS model the features in the multi-spectral images are estimated and classified. The estimated features are optimized with the stimulated annealing model for the estimation and classification of patterns in rural images. Based on the computed and estimated features the LSTM-based deep learning model is implemented for the pattern classification in the rural area. By utilizing multi-spectral data, the model captures a broader range of information, enabling a more comprehensive analysis of rural landscapes. Specifically, the GSA-MS model optimizes the extracted features using a simulated annealing algorithm, which iteratively refines the feature set to improve pattern estimation and classification accuracy in rural images. Additionally, the paper proposes the integration of a Long Short-Term Memory (LSTM) based deep learning model for further enhancing pattern classification accuracy in rural areas. Simulation results demonstrated that the proposed GSA-MS model achieves a higher classification accuracy of 99% for the estimation of patterns in the images with a minimal loss of 0.09.

Hidden Feature Weighted Deep Ranking Model (Hfwdr): A Novel Deep Learning Approach to Investigate the Nuanced Aesthetic Value of the Elderly Furniture Design & Cultural Identity

Jing Lu — 2024-03-26

The aesthetic value of elderly furniture design transcends mere functionality, embodying a rich tapestry of cultural heritage and historical significance. Rooted in traditional craftsmanship and informed by generations of cultural evolution, elderly furniture design carries with it a sense of timelessness and authenticity. Cultural identity and the evolution of the times are intertwined forces that shape societies, influencing everything from art and architecture to social norms and values. Cultural identity encompasses the unique customs, traditions, and beliefs that define a community or group, providing a sense of belonging and continuity across generations. This study investigates the aesthetic value of elderly furniture design, exploring its connection to cultural identity and the evolving socio-cultural landscape. By employing the Hidden Feature Weighted Deep Ranking Model (HFWDR), a novel deep learning approach, the research delves into the nuanced features of elderly furniture designs that resonate with cultural heritage and contemporary sensibilities. Through an analysis of design elements, material choices, and cultural motifs, the study uncovers the intrinsic relationship between furniture aesthetics and cultural identity, shedding light on how design evolves over time while retaining cultural authenticity. The HFWDR model, with its ability to capture hidden features and prioritize their significance in ranking, offers a comprehensive framework for evaluating and understanding the aesthetic evolution of elderly furniture design within the context of changing cultural dynamics. the HFWDR model assigned numerical values to hidden features such as symmetry, material quality, and historical relevance, with scores ranging from 0 to 100, indicating the degree of importance in determining the aesthetic value of elderly furniture designs.

Real-Time Optimization and Fault Diagnosis Algorithms for State Event Analysis in Elevator Group Control Systems

Jie Yu — 2024-03-26

Real-time optimization and fault diagnosis algorithms are essential components of elevator group control systems, ensuring efficient operation and timely detection of malfunctions. These algorithms continuously analyze the state events within elevator systems, such as passenger demand, elevator positions, and system performance metrics, to optimize elevator dispatching and minimize passenger waiting times. Moreover, they employ fault diagnosis techniques to detect anomalies or failures in elevator components, such as motors, sensors, or control systems, enabling prompt maintenance interventions and ensuring system reliability. By combining real-time optimization with fault diagnosis, elevator group control systems can enhance operational efficiency, passenger safety, and overall system performance, contributing to a seamless and reliable vertical transportation experience. This paper presented an approach to real-time optimization and fault diagnosis in elevator group control systems, integrating Trickle Transition State Event Analysis (TTSEA) techniques. Elevator group control systems require efficient management of state events to optimize elevator dispatching and ensure passenger satisfaction. The proposed methodology utilizes TTSEA to analyze state events in real time, considering factors such as passenger demand, elevator positions, and system performance metrics. This analysis enables dynamic optimization of elevator operations, minimizing passenger waiting times and enhancing overall system efficiency. Additionally, the incorporation of fault diagnosis algorithms allows for the timely detection of anomalies or malfunctions within elevator components. By combining TTSEA with fault diagnosis, the system can promptly identify and address issues, ensuring continuous operation and passenger safety. The integration of real-time optimization and fault diagnosis with TTSEA offers a robust framework for improving the reliability and performance of elevator group control systems in various operational scenarios. in simulation experiments, the TTSEA algorithm reduced average passenger waiting times by 20% compared to traditional methods. Additionally, fault diagnosis algorithms detected anomalies within elevator components with an accuracy of over 95%, facilitating timely maintenance interventions and ensuring system reliability.

Bibliometric Cluster Analysis and Classification Algorithm for Questioning Effectiveness in Elementary School Classrooms

Rong Wang — 2024-03-26

Bibliometric cluster analysis and classification algorithms offer a systematic approach to evaluating the effectiveness of questioning techniques in elementary school classrooms. By analyzing a vast array of academic literature and educational resources, bibliometric methods identify key themes, trends, and patterns related to questioning effectiveness. This analysis enables the classification of questioning techniques based on their impact on student engagement, comprehension, and critical thinking skills. By leveraging advanced algorithms, educators can gain valuable insights into the most effective questioning strategies for enhancing learning outcomes in elementary classrooms. This paper presented a novel approach to assessing the effectiveness of questioning techniques in elementary school classrooms through Bibliometric Cluster Analysis and a Classification Algorithm, integrating Centroid Clustering Deep Learning (CCDL). By analyzing a diverse range of scholarly literature and educational resources, the bibliometric analysis identifies key themes, trends, and patterns related to questioning effectiveness. Subsequently, CCDL is employed to cluster and classify questioning techniques based on their impact on student engagement, comprehension, and critical thinking skills. Through this integrated approach, educators gain valuable insights into the most effective questioning strategies for enhancing learning outcomes in elementary classrooms. the algorithm identifies three primary clusters of questioning techniques: low-engagement (average student participation rate below 30%), moderate-engagement (30-60% participation rate), and high-engagement (above 60% participation rate).

Gaussian Markov Chain Deep Neural Network Investigation for College Graduates' Initial Employment and Long-Term Career Development from an Economic Perspective

Xinyue Zhang — 2024-03-26

The initial employment and long-term career development of college graduates are critical topics from an economic perspective, with implications for both individuals and society as a whole. Examining graduates' entry into the labor market provides insights into broader economic trends, such as job availability, wage levels, and skill demands. Several issues affect the initial employment and long-term career development of college graduates. These include mismatched skills and job requirements, resulting in underemployment or unemployment among graduates. This study examines the initial employment and long-term career development of college graduates from an economic perspective, employing the Gaussian Markov Chain Deep Neural Network (GMC-DNN) for analysis. By integrating economic theory with advanced machine learning techniques, the research aims to elucidate the complex dynamics underlying graduates' labor market outcomes and career trajectories. Through the GMC-DNN model, which combines the capabilities of Gaussian Markov Chains for time series analysis and Deep Neural Networks for nonlinear pattern recognition, the study explores factors influencing graduates' employment transitions, wage growth, and career advancement prospects over time. Additionally, the model provides insights into the impact of economic factors, such as GDP growth, industry trends, and labor market conditions, on graduates' career trajectories. Simulation results demonstrated that the average starting salary for college graduates is found to be $50,000, with variations across fields of study and geographic regions. Furthermore, the GMC-DNN model predicts a median wage growth rate of 3% per year for the first five years of employment, with graduates in STEM fields experiencing higher wage growth rates compared to those in the humanities. Additionally, the simulation reveals that economic recessions lead to temporary setbacks in wage growth, with an average decrease of 2% observed during recessionary periods.

Optimization of Security Algorithms for Digital Authentication and Electronic Signatures in International Electronic Commerce Regulations

Zixiao Lu — 2024-03-26

Digital authentication and electronic signatures play a crucial role in international electronic commerce laws, facilitating secure and legally binding transactions across borders. These technologies enable parties to authenticate the identity of individuals or entities involved in electronic transactions and provide assurance regarding the integrity and non-repudiation of electronic documents. International laws and regulations, such as the United Nations Commission on International Trade Law (UNCITRAL) Model Law on Electronic Commerce, provide a framework for the recognition and enforcement of electronic signatures and authentication methods across jurisdictions. This paper focuses on optimizing security algorithms for digital authentication and electronic signatures in the context of international electronic commerce laws, utilizing the Homomorphic Elliptical Cryptograph Deep Neural network (HMEC-DNN). In an increasingly digitalized global economy, ensuring the security and integrity of electronic transactions is paramount. The HMEC-DNN framework combines the robust cryptographic properties of homomorphic elliptical cryptograph algorithms with the powerful pattern recognition capabilities of deep neural networks. Through rigorous optimization and training, the HMEC-DNN model enhances the efficiency and reliability of digital authentication and electronic signature processes, mitigating risks associated with identity fraud, data breaches, and transaction tampering. The results demonstrate significant improvements in security and efficiency compared to conventional methods. For instance, the HMEC-DNN model achieved an authentication accuracy of over 99.5% in verifying electronic signatures, ensuring the integrity and authenticity of electronic documents. Additionally, the framework reduced computational overhead by 30%, enabling faster transaction processing times and enhancing user experience.

Crop Yield Maximization Using an IoT-Based Smart Decision

Amita Shukla, Krishna Kant Agrawal — 2024-03-26

This paper conducted a comprehensive analysis of the integration between sensor technologies and machine learning algorithms in terms of crop yield prediction for precision agriculture. Appreciating the role of precision yield estimate in mitigating global food challenges, this paper discusses various sensor technologies including NPK sensors among others; their strengths and weaknesses are highlighted. An in-depth analysis of machine learning algorithms such as Decision Trees, Naïve Bayes, Support Vector Machines, K-Nearest Neighbors and Ensemble Learning reveal their comparative performances with regards to adopting them into agricultural practices. In addition, the use of Multiple Linear Regression for planning rainfall enables an interdisciplinary approach to precision agriculture based on both soil characteristics and climatic conditions. The discussion covers the emerging trends, patterns and gaps in previous research evidence on this topic along with possible implications for future studies or concrete implementation. Through identifying the challenges and limitations, including periodic sensor calibration as well as algorithm interpretability furthered by the review our complex reality of precision farming.

Indian Stock Market Sell and Buy Indication using Technical Indicators and Enhanced Bidirectional Long Short-Term Memory

Bhagyashree Pathak, Snehlata Barade — 2024-03-26

This study introduces an innovative approach to signal generation for sell and buy decisions in the Indian Stock Market, leveraging Novel Technical Indicators and an Enhanced Bidirectional Long Short-Term Memory (BiLSTM) model. We evaluated various machine learning models, including Random Forest, Gradient Boosting, XGBoost, DenseNet, CNN-BiLSTM, LSTM, BiLSTM, and DANN, on their predictive performance using metrics such as MAPE, MAE, and computation time. Our proposed BiLSTM model, optimized with novel technical indicators, demonstrated superior performance with the lowest MAPE and competitive MAE, while maintaining a rapid computation time. These results highlight the efficacy of BiLSTM models in handling the sequential nature of stock data and the advantage of novel technical indicators in capturing intricate market trends. The proposed system holds the potential to revolutionize decision-making processes for traders and investors by providing highly accurate, real-time market predictions. The comparative analysis across diverse machine learning techniques showed that the proposed method significantly surpasses conventional models like Random Forest, Gradient Boosting, and XGBoost in terms of accuracy and efficiency. It achieved a remarkable reduction in Mean Absolute Percentage Error (MAPE) to nearly 0.03%, drastically lower Mean Absolute Error (MAE) at 10.45, and exhibited the fastest execution speed at 0.984 ms, highlighting its substantial advancement over existing approaches.

Analysis of SNR of Physical Layer of 802.16e WiMAX under AWGN Channel and Doppler impact for Rician Blurring Channel with various Digital Modulation Techniques

Mahesh Pasari, Santosh Pawar — 2024-03-26

The drive to provide high-speed, cost-effective, and broadly accessible broadband wireless access (BWA) technologies has been a cornerstone in the evolution of the 802.16 standards, known as WiMAX. This standard outlines the air interface specifications for fixed BWA systems, operating within the 10-66 GHz frequency range. A significant body of research, including the model discussed in this context, has been devoted to evaluating the performance of the WiMAX Physical Layer (PHY) across various channel conditions, employing different digital modulation techniques. This analytical approach specifically focuses on measuring the system's efficacy by observing the impact on packet and bit loss rates under the IEEE 802.16 standard. Through such in-depth analysis, the model aims to shed light on the potential and limitations of WiMAX technology in delivering robust and efficient wireless broadband services.

A Model to Automate the Development of Computer Science Curriculum Syllabi

2024-03-26

Creating curricula and syllabuses is a crucial aspect of education. How well students are trained in computer science determines how good the students will be in the subject.A given course's syllabus is determined by several factors, including the objectives of the course, the resources available, the amount of time allotted, feedback from previous students, employers, and alumni, the aptitude of the learners, etc. Before developing a syllabus, course experts use a manual method that takes into account some variables and updates the syllabus as needed. Automation of the syllabus creation process is important because the needs of the software sector are always changing.This paper put out a model to make the process of creating syllabuses easier. The capability to design, modify, and store curricula will be offered by the model. Our model takes into account two factors: input from the industry and open-source course curricula from different universities. It will provide recommendations to the person who prepared the syllabus regarding its content based on these factors. The syllabus creator will finalize the contents by considering the suggestions given by the model and other attributes upon which it depends.

An Empirical Evaluation of Clustering Techniques for the Oral Cancer Prediction

S. Sivakumar, T. Kamalakannan — 2024-03-26

Data Mining is now widely used in healthcare applications to predict various cancers such as breast, kidney, thyroid, Colorectal, ovarian and many others. Clustering in Data Mining offers a solution for determining the prediction of Oral Cancer. This research explores K-means algorithm and introduces a new novel algorithm, the Kohonen map with K-means (Koho K-means). The experimental findings are based on 3004 oral cancer datasets, focusing on the time complexity and accuracy of the algorithms. The comparative study is then conducted with varying cluster points. The experimental results prove that Koho K-means outperforms K-means in predicting oral cancer, particularly in terms of accuracy.

A Hybrid Probabilistic Graph and Link Prediction Model for Complex Social Networking Data

2024-03-26

¹ Research scholar, Department of computer science and engineering, Koneru Lakshmaiah education foundation, Vaddeshwaram, AP, India

²Assistant Professor, Koneru Lakshmaiah Education Foundation, Vaddeshwaram, AP, India

³P.A. College of Engineering Mangalore, Affiliated to Visvesvaraya Technological University Belgum.

⁴Professor, Institute of Aeronautical Engineering, Hyderabad.

In this complex datasets of social networking, the possibility based graph community identification acts a prominent role. As many of the traditional models are intricate in estimating the novel link prediction type by utilizing benchmark graph community grouping measures. Besides conventional clustering measures utilize measures of nearest-neighbour regardless of contextual identicality for estimating the association amongst diversified nodes of graph. For optimizing contextual clustering of node & estimation of link, the hybrid scalable measure has been projected for clustering the community on intricate networks. Hence, in this research, the hybrid clustering graph & link prediction models have been projected on intricate social-networking dataset for effective patterns of decision-making. The simulation outcomes assist that projected contextual probabilistic-graph-clustering & link estimation model is having better effectiveness when compared to traditional approaches on intricate datasets of social-networking.

Prioritized Compressed Data Acquisition Framework for Securing the Data Integrity in the Medical Wireless Sensor Networks

B. Naresh Kumar, M. Srinivas — 2024-03-26

MWSNs are essential in enhancing healthcare systems by delivering efficient and effective solutions. These allow elderly and disabled individuals to live independently while ensuring their well-being and safety. In hospitals, the data collected by the sensors deployed in these networks can be analyzed in real time. To ensure the safety and confidentiality of the data collected by MWSNs, it is important that the information is properly aggregated and transmitted to the appropriate central servers. This process should be performed with the utmost security to avoid unauthorized access and manipulation of the data. This paper proposes a prioritized compressed data acquisition framework that is designed to increase the efficacy of gathering various health information types. The framework utilizes a sampling technique known as compressed sensing to reduce the transmission overhead and power consumption. Data is then encrypted, and its integrity is protected through a cryptographic hash protocol. The priority of the information is then considered, and the transmission is performed according to the usual method. The security aspects of the proposed framework are analyzed to ensure that the data collected by the MWSNs is protected. In addition, we performed an experiment on the system's performance. The findings of the study indicate that the proposed system can help improve the quality of healthcare by allowing more accurate and timely information.

Simulation of Artificial Intelligence based Robotic Arm for Patients with Upper Limb Amputations

Rajkumar Chougale, Vinay Mandlik, Asit Kittur, Vikas Patil, Ranjeet Suryawanshi — 2024-03-26

A myo-electric controlled prosthetic extremity is a prosthetic limb that seems to be controlled but is really controlled by electrical signals that the muscle structure itself automatically delivers. Electromyography is a novel method for recording and analysing electrical activity generated by muscles. Computerized reasoning and machine learning are particularly impressive in the mechanical and biological sciences. The purpose of this work is to apply artificial intelligence to predict and comprehend prosthetic hand movements using muscle training data. This idea already exists in the mechanical world, but it is prohibitively expensive and unavailable to non-industrialized countries. As a result, the primary goal of our research is to develop the much more precise intelligent bionic hand. In this research, also used MyoWare Muscle Sensor data, a tool that continually analyses information from eight sensors are also employed. Artificial intelligence and the informative index were used to anticipate finger, finger-close, round grip, and satisfactory-squeeze impulses. We It is next applied a few Artificial intellogence computations to the statistics verified with the 8-terminal superficial Electromyography MyoWare Strength Detector, including K-closest Neighbor (KNN), Support Vector Machine (SVM), and a mixture of SVM and KNN. In this research it is further characterised the four demonstrations of our prosthetic hand with a unceasing test accuracy of 98.33 percent by merging SVM and KNN. This report also includes a 3D visualisation of the robotic finger and its control strategy using Autodesk 3D's Max software design, an EMG MyoWare Muscle Sensor, Artificial intelligence.

BERT Model Based Identification and Classification of Web Vulnerabilities Using Deep Learning Approach

Manjunatha K. M., M. Kempanna, Pushpa G., Rangaswamy M. G. — 2024-03-26

In recent years, researchers have been focused upon machine learning and machine language based models to predict and identify effects of their researches. In this research the vulnerabilities in web, using the machine learning model BERT (Bidirectional Encoder Representations from Transformers) with additional layers have been attempted. The datasets used for the model’s prediction and classification are SQLInjection (SQLI) (namely: attacks and benign) and Cross Site Scripting (XSS) datasets respectively. The developed BERT model predicts the vulnerabilities in the data and classifies them accordingly. The loss is estimated through cross entropy loss technique. The performance of the model is evaluated through metric evaluation method namely binary accuracy. The analyses and findings shows that the developed advanced BERT obtained higher accuracy (SQLI with 98% and XSS with 97% accuracies respectively), than the standard BERT model (SQLI with 87% and XSS with 84% accuracies respectively). The research concludes stating that an increased BERT layers based model performs significantly with higher accuracy in classification than the standard BERT as a transformer model.

Naïve Bayes Classification of Sentiments on Subset using Tweets-during Covid-19

V. Geetha, N. Sujatha, Latha Narayanan Valli — 2024-03-26

Now a day, Social Media create a platform for almost all people for sharing and communicating with one another. Most of the business people and the organizations avail the social media conversation for their product promotion or predicting people behavior. The popular Social Media Networks are Facebook, Twitter, LinkedIn of Social Networks, Instagram, YouTube of Media Sharing Networks, Whatsapp, Pinterest and tripAdvisor of Consumer Review Networks. A Text Mining tool, Sentiment Analysis can help us to predict and classify the susceptible text used in the social media conversation. Even though having lots of advantages, unfortunately we have many risks in the usage of the social media content. Any individual must follow the rules and regulations for accessing the content in the social media networks. The objective of this research paper is to understand the various techniques involved in Sentiment Analysis process and choose to apply naïve bayes machine learning model in the subset level using twitter data to classify the sentiments of people in a best way.

An Adapted Moth Search with Convolutional Neural Network with Replicator Neuron-Based Leaf Disease Detection

Majed Aborokbah — 2024-03-26

Crop quality and yield can be significantly impacted by plant diseases, and even though plants may be examined for indicators of illness by trained biologists or farmers, this is typically an inexact and labor-intensive process. This study employs IoT and AI-based monitoring strategies to design and develop a smart method for classifying leaf illnesses. So as to measure the effectiveness of these two approaches, simulation results are compared in this work. In the first section, the data of photos of plants from the Plant Village data set augmented using a Hybrid CNN (Convolutional Neural Network) with RNN (Replicator Neural Network) and named as HCRNN, and deep features mined from these images. So as to enhance the precision of the segmentation procedure, the plant images undergo preliminary processing with an adaptive kaun filter. Next, a Glowworm Swarm Optimization based Clustering (GSOC) technique is used to isolate the plant region in the processed image. The HCRNN was then used to classify the plant disease based on the retrieved features. The projected method uses the Adapted Moth Search (AMS) Algorithm to fine-tune the CNN's (Convolutional Neural Network) hyperparameter in order to enhance its classification accuracy. Two branches of the model are used to learn from the T2- and Diffusion-weighted MRI data: one employs a ten-layer CNN After applying HCRNN to classify the relevant characteristics, and then assesses the quality of the classification in terms of precision, recall, and f-score. Extensive field testing indicates that the technique is useful in hot and humid environments and that it is more accurate than other categorization schemes at recognizing classes of disease in leaves.

Smart Agriculture: IoT and Machine Learning for Crop Monitoring and Precision Farming

2024-03-26

Through the implementation of cutting-edge technology like the Internet of Things (IoT) and machine learning, smart agriculture, which is also referred to as precision agriculture, is bringing about a revolution in the conventional agricultural techniques that have been used for generations. The purpose of this article is to present an overview of how the Internet of Things (IoT) and machine learning are utilised in crop monitoring and precision farming in order to improve production, maximise resource usage, and reduce environmental consequences. The Internet of Things (IoT) devices that are connected with sensors are placed throughout agricultural fields in order to collect real-time data on a variety of environmental characteristics. These parameters include soil moisture, temperature, humidity, and nutrient levels. These sensors are connected to one another over wireless networks, which enables the transmission of data to centralised cloud-based platforms for statistical analysis in a smooth manner. In order to recognise patterns, correlations, and anomalies in the data that has been collected, machine learning algorithms are applied to the subject matter. The development of predictive models allows for the forecasting of agricultural yields, outbreaks of pests and diseases, and the implementation of ideal irrigation schedules. In order to enable farmers to make educated decisions about irrigation, fertilisation, pesticide application, and crop management methods, decision support systems offer them with recommendations and alerts that may be put into action. The report also discusses the Internet of Things (IoT) and machine learning for crop monitoring. In addition to that, challenges associated with precision farming are discussed in this research.

Decentralized and Trustworthy Connectivity in IoT through Blockchain-Enabled Secure Data Sharing over Wireless Networks

2024-03-26

The proliferation of Internet of Things (IoT) devices has led to an exponential increase in data generated and shared over wireless networks, posing significant challenges in terms of security, privacy, and trustworthiness. This paper proposes a novel approach to address these challenges by leveraging blockchain technology for secure data sharing in IoT ecosystems. By decentralizing data storage and enabling tamper-resistant transaction records, blockchain provides a robust framework for ensuring the integrity, confidentiality, and accountability of IoT data. This paper outlines the key principles and mechanisms underlying blockchain-enabled secure data sharing in IoT networks, including cryptographic techniques, consensus algorithms, and smart contracts. Furthermore, it discusses the potential benefits and challenges of adopting blockchain in IoT deployments, as well as future research directions to overcome existing limitations and realize the full potential of decentralized and trustworthy connectivity in IoT environments.

Enhancing Financial Fraud Detection in Banking Systems: Integrating IoT, Deep Learning, and Big Data Analytics for Real-time Security

2024-03-26

Financial fraud poses a significant threat to banking systems, with increasingly sophisticated attacks targeting sensitive customer data and financial transactions. This paper proposes an innovative approach to enhance financial fraud detection in banking systems by integrating Internet of Things (IoT), deep learning, and big data analytics for real-time security. By leveraging IoT devices to gather real-time transaction data and user behavior patterns, coupled with advanced deep learning algorithms and big data analytics techniques, banks can detect and prevent fraudulent activities more effectively. This paper outlines the key principles and mechanisms underlying the integration of IoT, deep learning, and big data analytics in financial fraud detection. Furthermore, it discusses the potential benefits and challenges of adopting this approach, as well as future research directions to improve real-time security in banking systems.

A Novel Fuzzy C—Mean Based Segmentation Technique for Spinal Cord Tumors from MR Images

Alam N. Shaikh, Nisha A. Auti, B. K. Sarkar — 2024-03-26

Image processing plays a crucial role in extracting meaningful information from images to enhance their utility and effectiveness. Among various techniques, image segmentation stands out as an efficient method for extracting and isolating specific features within images. This research focuses on optimizing the Fuzzy C Means (FCM) algorithm for accurately identifying the axial and coronal planes in MRI brain images, considering both the algorithm's accuracy and computational efficiency.

The preprocessing phase involves converting MRI brain images from DICOM format to a standard image format. To enhance image quality, a Gaussian filter technique is applied to eliminate noise. Subsequently, the FCM algorithm is implemented to segment regions affected by brain tumours in MR images. The evaluation of algorithmic efficiency and accuracy involves comparing histogram values of images before and after segmentation with the cluster center values determined by the FCM algorithm.

The results provide insights into the algorithm's performance, with a focus on computational time as a key metric. By identifying the best fit of the FCM algorithm for both axial and coronal planes, this research contributes to advancing the field of image segmentation in the context of brain tumor detection. In conclusion, the study underscores the significance of FCM algorithm in accurately delineating tumor-affected regions in MRI brain images, thereby aiding in the diagnosis and treatment of brain tumors. The identified optimal parameters showcase the potential of FCM as a valuable tool in the realm of medical image analysis.

Scale-Invariant Feature Extraction for Skin Image Detection

Sami Hussein Ismael, Adel Al-Zebari, Shahab Wahab Kareem — 2024-03-26

In many applications, including dermatology, biometrics, and medical diagnostics, skin image detection is essential. Because of the differences in lighting, positions, and scales, it is difficult to identify skin regions in images. The article presents a new method for scale-invariant feature extraction-based skin image detection. The proposed strategy makes use of scale-invariant features to improve the skin image detection's resilience at various scales. Scale-invariant feature transform (SIFT) is used to extract key points from skin images, enabling the identification of unique patterns regardless of their size. The skin portions in the image are then reliably represented by using these key points. The incorporation of machine learning methods to improve the skin image recognition procedure is also explored in this research. A model is trained on a broad dataset of skin photos to enable the system to learn and adapt to different skin kinds, circumstances, and image scales. The evaluation's findings show how well the suggested scale-invariant feature extraction technique works to recognize skin images with reliability and accuracy.

Architecture Patterns Clustering using a Machine Learning Approach

Omar Al Huniti, Khawla Al-Tarawneh, Esra Alzaghoul, Fawaz Ahmad Alzaghoul — 2024-03-26

Architecture patterns are frequently employed in software development to address prevalent design challenges. The identification and classification of architecture patterns have become crucial in optimizing the design process due to the increasing complexity of software systems. Clustering has emerged as a widely adopted technique to categorize comparable entities. Recently, machine learning algorithms have been employed to automate and enhance the precision of clustering.

This study proposed using k-means clustering to group the architectural patterns like repository, client-server, broker, microkernel, publisher-subscriber, model view controller, REST, and space-based patterns together.

That was done on one of the benchmark dataset (Architectural Patterns dataset ) by using different Ks to perform the clustering, demonstrating the connections between architecture patterns. Then extract the related patterns and propose valid splitting for some patterns.

Masked Face Detection and Recognition System Using HOG Algorithm

Maryam Sarmad M. Ali, Fattah Alizade — 2024-03-26

The emergence of COVID-19 pandemic at the end of 2019 has introduced new problem to the face detection and recognition systems as faces are covered with masks that in return lead to reduction in the accuracy of face identification. For this reason, new adaptation to the previously available methods are becoming a necessary work in this area. In this paper, at first we are going to create a masked dataset to be used in our work that will mainly be divided into two parts, first to propose an algorithm for mask detection using Viola Jones algorithm, and second to suggest a face recognition algorithm that would use two methods namely (using pre-trained convolutional Neural Network CNN architectures Resnet-50 and Mobilenet, and building a customized CNN). The evaluation of this method was done on the dataset that we created at the beginning of our work based on the use of FEI dataset. The mask detection algorithm used has provided and accuracy of 79% whereas the face recognition methods has shown accuracy ranging between 75% and 99%.

Developing a Vehicle Monitoring and Tracking System using the Internet of Things (IoT)

Zina Balani, Naska Ismael Mustafa — 2024-03-26

Public transportation plays a vital role in daily life, serving as a key mode of transportation for individuals traveling between their homes, workplaces, and educational institutions. However, the inconvenience of extended waiting times for transportation often leads to time wastage. Many students frequently experience prolonged waiting periods at bus stops, eager to know the real-time location of buses and their estimated arrival times. To address these challenges, this research introduces the creation of an online system designed to visualize the routes taken by each bus during its journey. Integrated with Google Maps, this system enables students to easily access bus routes and schedules. The platform provides comprehensive details, including up-to-date information on the live location of buses displayed on the map interface. Additionally, the system furnishes supplementary information about bus drivers, encompassing their names, contact numbers, bus identifications, and the start and end times of their shifts. Users have the flexibility to access this data from any location, whether it's from their residence, workplace, or educational institution, using the web-based application and an internet connection. Moreover, a QR code scanning feature is available at bus stops, facilitating swift access to the desired information. By implementing this system, users, especially students, benefit from improved visibility into bus routes and schedules, enabling effective planning and minimizing wait times. The online platform delivers convenience and easy access, empowering users to make well-informed decisions regarding their journeys. The integration of real-time bus tracking, driver details, and QR code scanning significantly enhances the overall user experience, presenting a comprehensive solution for public transportation users.

CMOS Transceiver Epiretinal Vision Restoration Retina Chipset using Wireless Inductive Coupling

Hima Bindu Katikala, Telagathoti Pitchaiah, Gajula Ramana Murthy — 2024-03-26

In medical implants the inductive coupling is the most common topology used for power transfer. Because of design complexity a wireless power transfer is mostly adaptable in the medical implants. In this paper, for the data progressing inside the implant a complementary metal oxide semiconductor transceiver (CMOS TRx) on-chip prototype is proposed with an inductive coupling power transfer for retina application. The CMOS TRx is an integrated device contains the data modulators, configured inductive, demodulator & epiretinal electrode array used to generate an action potential inside the retina. The binary phase shift modulator with class-E power amplifier is designed to transfer the data within two carrier cycles at the transmitter side to maintain high modulation rate than other modulation techniques. The proposed inductive coupling tuned to be resonant at the frequency of 13.56MHz as same as bpsk carrier frequency with more parallel receiver resonance (ωp) by considering the retina permittivity.

Machine Learning Mastery: Leveraging Convolutional Neural Networks to Classify Skin Cancers as Benign or Malignant in the ISIC Database

Upendra Singh, Krupa Purohit, Chitralekha Dwivedi, Ritu Patidar, Sanjay Patidar — 2024-03-26

This research tackles the urgent need for enhanced precision in the detection of skin cancer, a common yet potentially deadly disease. Traditional diagnostic techniques frequently fall short in accuracy, prompting unnecessary and invasive medical interventions. Previous attempts to employ machine learning for distinguishing among different types of skin cancer have not been fully successful in achieving effective differentiation. To address these challenges, the study proposes an innovative approach utilizing Convolutional Neural Networks (CNN) for the autonomous identification of skin cancer. The designed CNN architecture incorporates three hidden layers, with the number of channels in each layer progressively increasing from 16 to 32, and then to 64. The model leverages the AdamW optimization algorithm with a learning rate set at 0.001, a choice that has proven to be highly effective. In evaluations conducted using the International Skin Imaging Collaboration (ISIC) dataset, which involved classifying skin lesions as either benign or malignant, the proposed CNN methodology demonstrated a remarkable accuracy rate of 96%. This level of precision indicates a significant advancement in the field of skin cancer diagnostics, highlighting the potential of CNN-based models to revolutionize the early detection and treatment of this condition.

Analysis of Exploratory Data For The Interactive Visualization Gap

Md Shahid Ahmad, Ravi Kumar, Md. Talib Ahmad, Niraj Kumar — 2024-03-26

Complexity is one of the defining features of data scaling. When it comes to big data and data integration, heterogeneous data is a major factor. Both are necessary, but enormous amounts of data processing and storage make it hard to see and understand big databases. Data extraction in a way that the human brain can understand is a major challenge in this data-driven era of exponential data growth. This study summarizes and offers a description of heterogeneous distributed storage, data visualization, and the difficulties associated with it, drawing on a variety of approaches from prior studies. In addition, we compare the findings of the examined research works and talk about the major change happening in the field of virtual reality huge data visualization.

Acquiring the Ability to Identifying Covid19 using Deep CNN from Impulse Noise in Chest X-Ray Pictures

2024-03-26

Utilizing CNNs, COVID19 is identified in X-ray pictures. Deep CNNs may have a harder time identifying things in noisy X-ray pictures. We provide a unique CNN technique that eliminates the need for preprocessing of noise in X-ray pictures by using adaptive convolution to enhance COVID19 detection. A CNN will therefore be more resistant to erratic noise. This method adds an adaptive convolution layer, an impulsive noise-map layer, and an adjustable scaling layer to the standard CNN architecture. Additionally, we employed a learning-to-augment technique with X-ray pictures that were noisy in order to enhance a deep CNN's generalization. The 2093 chest X-ray photos are divided into 1020 images showing a healthy image, 621 images showing pneumonia other than COVID-19, and 452 images showing COVID19. The architectures of pre-trained networks have been modified to increase their resilience to impulsive noise. Validation on noisy X-ray pictures showed that the proposed noise-robust layers and learning-to-augment strategy incorporated ResNet_50 led to 2% better classification accuracy than the present-day method..

Implementation of FinFET based 14T SRAM Memory Cell using Modified Lector Technique & Dual Threshold

Ramesh Gullapally, N. Siva Sankara Reddy, P. Chandra Sekhar — 2024-03-26

Static Random Access Memory (SRAM) can't become any smaller because to the sorts of materials and leakage testing used in today's large-scale silicon MOSFETs. This study uses a double edge and modified LECTOR method to create new tradeoffs in 14-semiconductor SRAM cells. Because of their outstanding transportation capabilities and the possibility for usage on large-scale processing and production, FinFETs are great replacements replacing outdated CMOS electronics. This test especially investigates a double edge value-based SRAM cell design. There are 14 semiconductor materials used in the design. According to the simulation's findings, FinFET-based circuits are more energy-efficient and reliable than scaled circuits. They are more susceptible to variances and flaws nevertheless. These findings suggest that in terms of power efficiency, FinFETs are far superior to CMOS.

An Ensemble Approach for Comprehensive Brain Tumour Detection Using MRI-Based Machine Learning Models

2024-03-26

In the realm of medical imaging, the rapid evolution of techniques and exponential growth of data have emphasised the significance of automatic and reliable tools for brain tumour detection. This project proposes a system designed to detect brain tumours utilising Magnetic Resonance Imaging (MRI) data. Two distinct models leveraging advanced machine learning algorithms, particularly Convolutional Neural Networks (CNNs), are developed using multiple datasets. The BraTS dataset is used for segmentation tasks, providing detailed information about specific brain tumour regions. Concurrently, the Br35H dataset is used for binary classification, distinguishing the presence or absence of tumours. Furthermore, the brain tumour dataset from Kaggle adds another dimension to this study, offering diverse data samples. The proposed system encompasses a two-step approach. First, a segmentation model is fine-tuned on the BraTS dataset to identify specific regions within brain scans. Subsequently, a classification model is trained using both the Br35H and Kaggle datasets. Ensemble learning techniques, involving an ensemble of CNN architectures such as ResNet and VGG, along with the exploration of ensemble methods like AdaBoost, are employed for effective classification.

Enhancing Session-Based Recommendations with GRU4Rec and ReChorus

Drashti Shrimal, Harshali Patil — 2024-03-26

Recommender systems have evolved from basic item-to-item recommendations to sophisticated, session-based algorithms. A pivotal model in this transition is GRU4Rec, which employs Recurrent Neural Networks (RNNS) for session-based recommendations. While GRU4Rec has shown marked improvements over traditional methods, its effective deployment necessitates a robust training frame- work. This paper leverages ReChorus, a PyTorch frame work designed for top-K recommendation with implicit feedback, to train the GRU4Rec model. ReChorus offers a streamlined model design process, high efficiency, and flexibility, making it well-suited for achieving state-of-the art metrics, specifically NDCG and Hit Rate. Empirical evaluations across multiple datasets confirm that this approach successfully matches existing state-of- the-art metrics in the field of Recommender Systems.

An Optimized Integer Representation through a Novel Numeric Encoding for Textual Data Compression

Kanak Pandit, Harshali Patil, Poonam Joshi,Tarunima Mukherjee — 2024-03-26

The objective of this paper is to introduce a new variable sized integer encoding technique for file compression. The paper aims to compare the performance of the proposed method with established codes like Elias Gamma, Elias Delta, and Golomb. The study also seeks to examine the impact of varying log base values on compression ratio and runtime efficiency. The proposed method utilizes radix conversion and the Burrows Wheeler Transform for file compression. Performance comparison is conducted on the Calgary corpus, which includes both text and binary files. Existing codes like Elias Gamma, Elias Delta, and Golomb are executed on the files before evaluating the proposed code. Graphs are used to analyze the impact of log base values on compression ratio, while runtime efficiency is assessed. The proposed compression code achieves varied compression ratios (1.67 to 1.87) at radix r=4, highlighting its effectiveness over existing algorithms. A non-linear relationship between the log base and compression ratio is observed, plateauing as the log base increases. Runtime varies among files, with 'bib1' at the longest time (6.41 seconds) and 'obj1' the shortest (0.09 seconds). A positive correlation exists between the number of data points (n) and runtime, while a negative correlation is seen between 'n' and compression ratio, indicating lower ratios for larger 'n' files. Comparing its performance with established codes provides a benchmark for evaluation. Analyzing compression ratio trends and runtime efficiency offers insights into the effectiveness of the proposed method, adding to its novelty.

Pictorama: Text based Image Editing using Diffusion Model

2024-03-26

This research aims to pioneer image modification through text, integrating natural language descriptions with advanced computer vision and NLP techniques. The primary objective is to bridge human language and image editing, empowering users to convey creative visions effortlessly, revolutionizing the field of image modification. The study employs stable diffusion models, leveraging PyTorch and Python. It builds on prior works like Imagic, LEDITS, and Instructpix2pix, integrating a novel Vector Quantized Diffusion (VQ-Diffusion) model. The model is trained on a dataset of 436 GB containing 3 features an input image, an editing instruction, and an output edited image. Test samples include real images subjected to diverse text prompts for image edits, with disentanglement properties explored. The approach combines text inversion and Box-Constrained Diffusion (BoxDiff) for personalized and conditional image synthesis. The research showcases that stable diffusion models exhibit disentanglement properties, enabling effective modifications without extensive fine-tuning. The introduced BoxDiff and VQ-Diffusion models demonstrate superior performance in spatially constrained and complex scene synthesis, outperforming traditional methods. We are able to observe greater quality in output images with good cohesiveness throughout the image. Runing the model with greater number of steps allows for upheaval in quality. Here we have used 100 steps for greater image quality. The effect of number of steps on the time taken for inference is also studied. Due to the large amount of video memory required for inferencing, we recommend a GPU with >11 GB of video memory. The study adds value by addressing biases, achieving higher speeds, and enhancing image quality, contributing to the evolving landscape of text-to-image synthesis. This research introduces novel approaches in disentanglement, spatially constrained synthesis, and rapid image generation, pushing the boundaries of text-to-image synthesis beyond existing limitations.

Enhancing User Recommendations through Context-Driven Natural Language Processing (NLP) and Strategic Feature Selection

Akanksha Pal, Abhishek Singh Rathore — 2024-03-26

The surge in popularity and significance of social networks in recent years is undeniable, with social networking sites experiencing an exponential increase in user engagement. These platforms enable users to connect with others, establishing friendships and facilitating communication. A notable trend among most social network websites is leveraging the social graph's proximity for recommending potential friends to users. The study in question introduces a user recommendation system that employs various algorithms to identify similarity factors among users, thereby enhancing the precision of friend suggestions. A key technique utilized in this system is feature selection, which effectively extracts pertinent information from both text and hypertext data sources. Among the various algorithms explored, the Context-Driven Network (CDN) stands out for delivering superior performance in generating user recommendations, indicating its effectiveness in harnessing contextual information to improve the relevance and quality of connections suggested on social networking sites.

Integrating Long Short-Term Memory and Reinforcement Learning in Federated Learning Frameworks for Energy-Efficient Signal Processing in UAV-Assisted Wireless Communication Networks

Mahesh Y. Sumthane, Kirti Saraswat — 2024-03-26

This paper presents a comprehensive study of signal processing algorithms designed for enhancing the energy efficiency of UAV-aided wireless communication networks. We explore a sequence of advanced machine learning techniques, each tailored to address specific challenges within the network. We begin by detailing the application of Long Short-Term Memory (LSTM) networks, which are adept at uncovering patterns in data with unknown objectives or constraints. Echo-State Networks (ESNs) are then introduced for their proficiency in sequence and pattern detection, essential for classification and regression prediction problems in signal processing. We further examine the role of Reinforcement Learning (RL) in actively engaging with prediction problems and NP-hard problems, leveraging a reward-based system to facilitate active learning. In addressing the critical concerns of data privacy and excessiveness, Federated Learning (FL) is proposed as a decentralized solution that promotes local training on UAVs, significantly reducing the need for data centralization. Through the methods outlined, we achieve a novel optimization framework that integrates the aforementioned techniques, commencing with the identification and mitigation of unwanted vehicles in the network, which is processed into a Data Traffic Matrix. This feeds into an LTE DIC algorithm based on correlation and culminates in an optimization process that considers specific network parameters 'P' and 'B'. The results, derived from the comparative analysis using the established techniques, indicate a significant improvement in network efficiency. The proposed framework demonstrates a marked enhancement in energy efficiency, with an observed improvement percentage over existing methods. This substantiates the efficacy of the integrated approach, suggesting that the application of machine learning algorithms can lead to superior performance in UAV-assisted networks, providing a significant step forward in the development of autonomous and efficient wireless communication systems.

Secure Model of Access Control for Cloud Computing using Key Generation Based Public Cyclic Key Generation Method

Ranjeet Osari Rahul Singhai — 2024-03-26

Cloud computing is a big platform of service-oriented applications over the internet. The primary access control of cloud services using login credentials for users. The growing rate of malicious software breaks the security credentials of users and theft data, and blocks the services. To prevent security threats, cloud service providers and NIST design various access control using cryptography algorithms. However, the role-based access control mechanism has limitations and breaks the security bridge between users and service providers. This paper proposed key generation-based access control methods for accessing services and data over cloud computing. The proposed key generation approach is a public key generation algorithm, a cyclic key generation algorithm. The proposed key generation methods are implemented in the Java RMI model and MYSQL database. The proposed algorithm compares with RSA based key authentication approach. The experimental results suggest that the proposed algorithm is better than the existing algorithm of access control of cloud computing.

Deep Learning-Infused Cascading Regression Approach to Predict the Academic Performance of Immigrant Students

Selvaprabu Jeganathan, Arun Raj Lakshminarayanan — 2024-03-26

The academic performance of immigrant students is governed by a diverse range of resources and contexts, including the families of the students, the immigrant communities from which the students originate, and the social and educational attitudes that are held toward immigrants in the countries in which the students are currently residing. The Program for International Student Assessment, is an educational research initiative that is used to assess the knowledge and skills of students who are 15 years old. In this paper, the performance of immigrant students is predicted using the PISA dataset. There are a total of 35 attributes present in the dataset. Among these, the proposed method chooses three attributes(maths, science and reading) as target variables for performance prediction. This research constitutes a novel cascading regression framework designed to accurately forecast academic performance. Sequentially integrating CatBoost Regressor, Bidirectional Recurrent Neural Network (Bi-RNN), and Random Forest Meta Regressor optimizes predictive accuracy. Initiated by the CatBoost Regressor, its outputs serve as inputs for a Bi-RNN model, exploiting bidirectional sequential information. The ensuing predictions from Bi-RNN inform a Random Forest Meta-Regressor, refining the final outcome. Evaluation metrics, comprising MAPE, RMSE, and R2, substantiate the superior accuracy of the cascading model. The cascading ensemble significantly outperformed all individual models, achieving a MAPE reduction of 3.74%, an RMSE reduction of 20.70%, and an R-squared increase of 0.96.This research highlights the efficacy of cascading ensemble techniques for predicting student performance with enhanced accuracy. The method being proposed demonstrates the capacity to capture both fixed and changing characteristics, which may result in enhanced interventions and educational decision-making.

Performance Optimization of Long-Haul Optical Transmission Link with Optical-OFDM

Jyoti Prashant Singh, Deepak Kumar Singh, B. B. Tiwari — 2024-03-26

This research focuses on improving communication system performance by implementing optical orthogonal frequency division multiplexing (OFDM). OFDM is utilized to extract data from high-capacity systems involved in extensive information transmission. Advanced computer programs and strategic methodologies are employed to develop a measurable and highly efficient communication system. The study delves into both theoretical and practical applications of these techniques. Using MATLAB simulations, the investigation into OFDM capabilities offers a robust platform for comprehensive analysis and evaluation. Researchers can easily adjust subcarrier separation, modulating schemes, and other key parameters to gauge their impact on system performance. This adaptability facilitates thorough exploration of various scenarios and optimization strategies. The findings of this study provide valuable insights into the effectiveness and feasibility of optical multiplexing with OFDM for high-capacity transmission in optical communication systems. Optical OFDM, through simulations, demonstrates outstanding spectrum utilization in the light domain compared to conventional radio-frequency (RF) OFDM, owing to its ability to tightly allocate subcarriers.

Cyber Security Framework for Manufacturing Industry with Robotic Process Automation integration

Murugappan K., T. Sree Kala — 2024-03-26

Cyber Security is becoming one of the business enablers in the current Industry. It is a necessity rather than added item as there were various security breaches that occurred in the past that brought many businesses to their knees. Especially it is a challenge in the Manufacturing industry because of the lack of security consideration in Operational Technology (OT) equipment and its surrounding processes. OT is a hardware and software-integrated platform that shall be used to monitor and control the physical device’s operation. A major portion of it is mechanical, and those that use digital controls have closed, proprietary protocols. However now the trend is moving towards smart technology, and convergence of Information Technology and OT is unavoidable. This paves the way for typical cyber-attacks in OT. It shall lead to critical infrastructure damages or malfunctioning, and those results in key services failure, and put human life at risk. Hence in this paper, we explored cyber security requirements in the manufacturing industry, robotic process automation integration and recommend a proposal to build an end-to-end Cyber Security framework.

Apple Disease Detection Using Convolutional Neural Networks

A. S. Lalitha, K. Nageswararao — 2024-03-26

Cultivation of a crop is a pertinent aspect in the agricultural sector. The infection of crops with a disease is one of the bottlenecks leading to reduction in the yield of the crop. This decreases the production rate and thereby creating problems in maintaining the crop throughout the year. Manual identification of the disease is laborious and time consuming. It is laborious in the sense that, it requires lot of expertise and monitoring and this problem is going to be much more, especially, when the crop is big enough to manage. In order to minimize the effort needed, A deep learning model has been developed to identify particular disease. Using this model it facilitates the farmer to detect the disease accurately in a real time. The model developed in this research is a Residual Neural Network model as it helps us in implementing the feature extraction and classification of a particular disease of a fruit. A database of 505 apples was considered and 385 apples were used for training the model and 120 apples were considered for testing. The proposed model has generated an accuracy of 78.76% with a loss value of 0.6818 in detecting the disease of an apple. This computer based model will enhance usability of the model in detecting the disease of a fruit in a feasible manner and at an early stage of the disease. This in turn enables the farmer to detect the disease at an early stage of the disease and can take the precautionary measures to cure the disease in a more effortless manner.

Pemantic Segmentation in Medical Imaging using U-Net Convolutional Neural Networks

2024-03-26

This research investigates the application of U-Net convolutional neural systems in the semantic division for restorative imaging, centering on brain tumor distinguishing proof and kidney tumor division. Four cutting-edge calculations, specifically U-Net, DeepLabv3, Mask R-CNN, and LinkNet, were comprehensively assessed. Through thorough experimentation on assorted therapeutic imaging datasets, the study uncovers that Veil R-CNN shows superior division exactness, accomplishing an amazing IoU of 91.3% and a Dice coefficient of 94.2%. Comparative investigations with related work illustrate the competitiveness of the proposed approach in comparison to state-of-the-art strategies. In addition, the investigation digs into computational productivity contemplations, generalization over modalities, and factual importance testing, advertising a comprehensive appraisal of algorithmic choices. Patterns such as privacy-preserving combined learning and the integration of worldly data in dynamic imaging modalities were also investigated. The results of this investigation not as it were contribute to the headways in the semantic division for therapeutic imaging but also give practical experiences for the improvement of precise and productive apparatuses with real-world clinical applications.

Image Super-Resolution with Deep Learning: Enhancing Visual Quality using SRCNN

2024-03-26

This research digs into the space of Image Super-Resolution, particularly centring on the assessment and comparison of four noticeable calculations: SRCNN, FSRCNN, ESPCN, and VDSR. The study utilizes a different dataset enveloping different spaces, from common scenes to restorative imaging and satellite applications. Quantitative measurements, counting Top Signal-to-Noise Proportion (PSNR) and Structural Similarity Index (SSIM), and nearby visual quality evaluations were utilized to gauge the execution of each algorithm. SRCNN developed as the frontrunner, showing the most elevated PSNR of 28.7 and a commendable SSIM of 0.89. FSRCNN and ESPCN were closely taken after with PSNR values of 27.9 and 28.3, and SSIM scores of 0.88 and 0.87, separately. VDSR illustrated competitive execution with a PSNR of 27.5 and an SSIM of 0.86. These quantitative results were complemented by visual quality appraisals, where SRCNN got the most elevated rating of 9.2, followed by FSRCNN (8.8), ESPCN (8.7), and VDSR (8.5). This research contributes to the continuous exchange of computer vision, emphasizing the qualities and trade-offs of each calculation and giving important bits of knowledge into their pertinence over differing picture super-resolution scenarios.

Genetic Algorithms and Machine Learning for Optimal Power Flow Solutions

2024-03-26

This research examines the application of Genetic Algorithms (GAs) and Machine Learning (ML) in tackling the Optimal Power Flow (OPF) issue inside control frameworks. The study points to playing down operational costs whereas assembly operational imperatives through the optimization of control factors. Tests were conducted comparing GAs, Particle Swarm Optimization (PSO), Bolster Vector Machines (SVM), and Neural Networks (NN). The comes about uncovered that GAs reliably outflanked other calculations, illustrating predominant merging speed and accomplishing lower add up to costs. The research contributes experiences into the viability of GAs in exploring the complex and non-convex arrangement space of the OPF issue. Comparative investigations with related works assist fortified the competitive execution of the proposed Genetic Algorithm approach. This consideration not only propels the understanding of control framework optimization but also gives profitable suggestions for the broader application of GAs and ML procedures over differing spaces. The research highlights the potential for crossover approaches and the integration of real-time information to enhance versatility and vigor within the setting of keen networks and feasible vitality systems.

Deep Learning Applications in Medical Image Analysis: U-Net for Diagnosis

2024-03-26

This research investigates the application of U-Net engineering in restorative image investigation for enhanced symptomatic capabilities. Leveraging a different dataset comprising MRI, CT scans, and X-rays, we methodically compare U-Net with conventional CNN, SegNet, and state-of-the-art DeepLabv3. The U-Net show showcases predominant execution, accomplishing a Dice coefficient of 0.85, an Intersection over Union (IoU) of 0.75, and a pixel exactness of 0.92. The incorporation of skip associations in U-Net demonstrates instrumental in protecting spatial data, driving more exact division comes about. Moreover, our examination amplifies to particular therapeutic conditions, illustrating U-Net's flexibility with a Dice coefficient of 0.87 for tumor division and 0.83 for organ outline. The results confirm U-Net as a vigorous and dependable instrument for exact medical picture division, with suggestions for improved demonstrative precision over different imaging modalities.

An Illustrative Review Cryptographic Algorithms for Scada Application in Networking Ntru for Security

2024-03-26

The computational applications and network protection capabilities of mobile and wireless networks have grown exponentially. Cyberattacks and other unlawful activity on commercial and personal networks have increased in recent years. Firewalls and security code fail to secure computer networks. Personal gadget users, company employees, and military personnel realise network defence is crucial. SCADA systems are commonly utilised in Critical Infrastructure Systems to autonomously monitor and control industrial activities. Security concerns are more likely in SCADA design due to its reliance on computers, networks, applications, and programmable controllers. The huge demand for computers among corporations and other organisations has led to the creation of several networks. Computer network attacks have increased in recent years. Cryptographic solutions of these requirements have performance issues in SCADA systems. NTRU, a faster and lighter public key technique for end-to-end security, is used in this research to improve SCADA security requirements.

Machine Learning Algorithms for IOT Services in Big Data and Cloud Computing

2024-03-26

The phrase "cloud computing" refers to a kind of data management system in which mobile devices are not used for either the processing nor the storing of user data. The Internet of Things (IoT), a brand-new technology that is only now entering its formative years, is also becoming more widespread in the networks and telecommunications sectors. The "modern" sector of wireless telecommunications networks is where the majority of the emphasis of application for the Internet of Things is now being directed. In the most recent part of our line of research, we investigated the relationships and interactions that exist between the many different entities and equipment that communicate across wireless networks. They need to achieve the goal that has been set for them as a group in order to make the atmosphere more conducive to the use of big data. This will help create a more favourable environment for the use of big data. This article discusses the Internet of Things (IoT) and Cloud Computing technologies, with a particular focus on the security challenges that each of these technologies has experienced. In the field of medicine, for instance, big data is being put to use in order to bring down the costs of treatment, anticipate the arrival of pandemics, prevent sickness, and carry out a variety of other related activities. This article provides a comprehensive introduction to the approach of big data analytics, which is crucial in a variety of fields of work and businesses. First, we will present a brief overview of the concept of big data, which refers to the quantity of data that is generated on a daily basis, as well as its characteristics and facets.

“Hybrid MAC Methodology for Improving the Qos in Fiber Wireless Network”

Prabhjot Kaur, Hardeep Singh Saini — 2024-03-26

The Fi-Wi (Fiber-Wireless) approach stands out as a crucial element in the realm of networks, demonstrating superiority over various technologies. With the exponential growth in Internet users, significant strides have been made in the evolution of Fi-Wi networking systems in recent years. This mechanism offers broader bandwidth and network stability, ensuring high-speed connectivity with "Anytime Anywhere" availability for end users. However, the escalating energy demand in networking systems poses a constraint on the network's lifespan, impacting transmission.

Over the years, researchers have proposed and tested various Media Access Control (MAC) protocols to address transmission and energy consumption issues. Despite these efforts, existing protocols have encountered challenges such as overheating, delays, throughput issues, and collisions. This research paper introduces a combination of Time Division Multiple Access (TDMA) and Carrier Sense Multiple Access (CSMA) to tackle associated challenges. The primary objective is to enhance throughput and reduce delays in Fi-Wi networks.

To achieve this goal, the study employs techniques that involve an Utilizing an Orthogonal Frequency Division Multiplexing (OFDM) modulator, a free-space optical (FSO) communication channel, an OFDM demodulator, and Opti-system for the analysis and enhancement of received signals, our study demonstrates that the proposed MAC protocol surpasses conventional MAC protocols in terms of delay, data throughput, and transmission efficiency.

Artificial Neural Networks (ANNs) used for change detection in remotely sensed images

2024-03-26

This paper examines the application of semi-supervised Artificial Neural Networks (ANNs) in the change detection of remotely sensed images. Relying on the analysis of multi –temporal satellite images to detect altercations caused by natural or human activities is crucial for change detection for monitoring environmental changes and urban expansion. Recent advancements in Artificial Intelligence (AI) particularly semi-supervised ANNs, have significantly improved the accuracy and efficiency of change detection processes. This review highlights various methodologies and techniques employed in the field, including the integration of Convolutional Neural Networks (CNNs) and Graph Convolutional Networks (GCNs) for enhanced feature extraction and classification. The paper discusses the application of these methods across different scenarios such as agricultural yield prediction, urban growth monitoring and environmental surveillance underlining the importance of ANNs in advancing remote sensing capabilities.

Machine Learning and Ai in Marketing–Connecting Computing Power to Human Insights

2024-03-26

Researchers' interest in artificial intelligence (AI) agents that are driven by machine learning methods has been piqued as a result of the rapid changes that these technologies are creating in the marketing sector. In the framework of this article, we investigate and argue in favour of the use of methods related to machine learning to marketing research. We provide a comprehensive overview of the common aims and methodologies of machine learning and compare them to the traditional statistical and econometric approaches that are employed by marketing professionals. In this research, we claim that machine learning approaches can analyse vast volumes of unstructured data, make accurate predictions, and have model structures that are adaptable. In addition to being difficult to understand, these methodologies are also unclear with reference to the models. We provide scalable and automated decision support capabilities, which are essential for business managers. We investigate the most important business trends and practices that are being driven by AI, as well as academic marketing research that combines machine learning approaches. Most significantly, we provide both a detailed plan for further research as well as an extensive conceptual framework.

Artificial Intelligence and Machine Learning as Business Tools: A Framework for Diagnosing Value Destruction Potential

2024-03-26

Machine learning (ML) and artificial intelligence (AI) have the ability to save expenses and increase the efficacy of corporate operations. On the other hand, they also have the capacity to devalue a company's assets, which may sometimes have extremely catastrophic effects. It's possible that some managers won't accept new technologies because they can't fully understand and effectively manage the risks associated with doing so. This will prevent them from realising their maximum potential. The findings of this study provide a fresh paradigm for detecting and limiting the value-reducing potential of artificial intelligence and machine learning for businesses. In addition to outlining the components of an AI solution, this research also recommends this paradigm. The paradigm might be used to map the components of an artificial intelligence system. The concepts of value-generation process and content are then used to illustrate how the aforementioned dangers have the potential to obstruct the creation of value or even result in the loss of that value. In the interest of shedding some light on the topic of the commercial activation of artificial intelligence, this study does an in-depth and careful examination of the existing body of literature on the topic. In addition to that, a clear and succinct explanation of what constitutes artificial intelligence at the present time will be provided. The Implications, Applications, and Methods model (also known as the IAM model) has uncovered a total of six topics that are associated with these three primary topics of discussion. It is possible that academics and practitioners will find our study beneficial in that it provides an overview of the body of knowledge and research agenda. This will allow for artificial intelligence to be used as a strong facilitator in the process of producing business value.

A Comprehensive Multimodal Approach to Assessing Sentimental Intensity and Subjectivity using Unified MSE Model

Mohd Usman Khan, Faiyaz Ahamad — 2024-03-26

In the dynamic realm of multimodal learning, where representation Learning serves as a pivotal key, our research introduces a groundbreaking approach to understanding sentiment and subjectivity in audio and text. Illustration from self-supervised learning, we've innovatively combined multi-modal and Unified--modal tasks, emphasizing the crucial aspects of consistency and distinctiveness. Our training techniques, likened to the art of fine-tuning an instrument, harmonize the learning process, prioritizing samples with distinctive supervisions. Addressing the pressing need for robust datasets and methodologies in combinational text and audio sentiment analysis, we offer the dataset for Multi-modal sentiment intensity assessment at the Opinion Level (MOSI). This meticulously annotated corpus offers insights into subjectivity, sentiment intensity, text features, and audio nuances, setting a benchmark for future research. Our method not only excels in generating Unified-modal supervisions but also stands resilient against benchmarks like MOSI and MOSEI, even competing human curated annotations on the challenging datasets. This pioneering work paves the way for deeper explorations and applications in the burgeoning field of sentiment analysis.

Fuzzy Intrusion Detection Method and Zero-Knowledge Authentication for Internet of Things Networks

2024-03-26

One of the encouraging trends that has contributed to the exponential rise in human progress over the last decade is the Internet of Things (IoT). Interconnection of physical objects for the purpose of data exchange is the next frontier of the internet, known as the Internet of Things (IoT). Everything from household appliances to cars to buildings to animals is part of the Internet of Things (IoT). The Internet of Things (IoT) has become the de facto standard because to its many useful uses in business, medicine, agriculture, and other fields. The Internet of Things (IoT) integrates wireless, pervasive, and ubiquitous technology to solve problems. Things with sensors implanted in them and linked over the internet make it up. Data is collected and shared by these networked devices. Many applications rely on the monitoring and analysis of data originating from heterogeneous devices. Fuzzy logic is used in this study to create a new, lightweight intrusion detection system (IDS) for Internet of Things (IoT) applications based on the MQTT protocol. Using fuzzy variables, the IDS detects network irregularities. In conclusion, this study offers a fresh security framework to solve the problems with current algorithms in an Internet of Things setting. This study also provides an application layer security that smart environments may use to avoid DoS attacks

Identifying Fake News on ISOT Data using Stemming Method with a Subdomain of AI Algorithms

2024-03-26

Nowadays, social media platforms have played a significant role in disseminating information throughout the world without any hindrance. Some people take this opportunity to propagate fake news in order to make money, by damaging the reputations of others. To tackle this issue, we proposed a methodology for detecting fake news on social media. This methodology extracts a feature of TF-IDF using N grams and Word2Vec in two ways i) with stemming method ii) without stemming method. Both of the process is performed and they fed an into supervised machine learning algorithms (ML) such as logistic regression (LR), random forest (RF), support vector machine (SVM), gradient boosting (Grad), adaptive boosting (Adaboost), and stochastic gradient descent (SGD) to detect a fake information. Evaluation shows that the unigram gives a better result with random forest when compared to the bigram and trigram. All classification algorithms were outperformed by Trigram. Unigram is more exact both with and without a stemming method. Word2vec has lower accuracy to detect fake information in the given dataset.

Machine Learning for Quantum Computing Bridging the Gap between AI and Quantum Algorithms

2024-03-26

This study explains how machine learning techniques are applied to enhance quantum algorithms and examines the interplay between machine learning and quantum computing. It explores quantum data analysis, quantum machine learning, and hybrid quantum-classical techniques, emphasizing their contributions to bridging the gap between artificial intelligence and quantum algorithms. Additionally, it analyzes how quantum data production, quantum-assisted optimization, and quantum neural networks could influence the direction of AI-quantum integration in the future.

Human Activity Detection using Profound Learning with Improved Convolutional Neural Networks

S. Anthonisamy, P. Prabhu — 2024-03-26

Human Activity recognition (HAR) is an interesting area of research mainly due to the availability of low cost sensors and accelerometers live streaming of data and advances in technology. HARs involve identifying various human activities such as walking, running, sitting, sleeping, standing, showering, cooking, driving, opening the door, abnormal activities, etc. are recognized. The data can be collected from wearable sensors or accelerometer. HARs can be extensively used in medical diagnostics for keeping track of elderly people, HARs approaches analyze data acquired from sensing devices, including vision and embedded sensors. HARs are assistive technologies mainly used for taking care of elders in healthcare. Approaches of HARs attempt to predict people’s movements often indoors and based on sensor data like accelerometers of smart phones. In terms of classifications, HARs are challenging tasks as they involve time series data where Deep Learning Techniques (DLTs) like CNNs (Convolution Neural Networks) have the ability to correctly engineer features from these raw data while building their learning models. This paper proposes Human Activity Detections using Profound Learning (HADPL) based on CNNs which detects HARs from captured accelerometer data. HADPL was tested on WISDM_Act_v1.1 dataset and evaluated for its performances in terms of precisions, accuracies, recalls and F1-scores where it achieved a decent level of accuracy by scoring up to 95 percent. The proposed technique can be implemented for monitoring elderly people based on captured or stored HAR data.

Intelligent Advanced Model Implementation of Green Financing Concept in the Financial Monitoring System for Enterprises Activity based on Sustainable Development

2024-03-26

According to the information that was presented in the article, financial monitoring is used as a critical control point observation tool to minimise the threat to the nation's overall financial stability. It is explained how the company monitoring system is compatible with the goals of a sustainable and environmentally friendly economy. Internal financial monitoring is required in order to put green finance initiatives into action, and it is also used to evaluate how well these efforts are being carried out. According to the reports, the organization's financial monitoring system acts as a tool for accomplishing the goals of sustainable development and a green economy through resource efficiency, the institutions of production, financial resources, and human resources. This is the case because of how the system was designed. This assertion is backed up by the idea that the company's financial monitoring system acts as a tool for the purpose of accomplishing the aforementioned goals. It is essential to underline that the State Financial Monitoring Service has tight linkages to both the national and global sustainable development plan and that, in its fight against financial crime, it takes into account international norms. Both of these points are crucial to emphasise. The term "green financing" refers to an emerging subfield within the field of finance that is concerned with the question of how to strike a balance between the competing goals of increasing economic activity and increasing environmental protection. It is essential to place equal focus on each of these features. In addition to establishing the function of internal and governmental financial monitoring in green finance, the essay underlines the need to investigate the investment environment for the green economy. This requirement comes after the article outlined the importance of internal and governmental economic growth in green finance. The essay also underlines how essential it is to take into account the business climate while pursuing a green economy.

Machine Learning for Alzheimer's Disease Detection and Categorization in Brain Images

2024-03-26

Alzheimer's disease (AD) is a devastating form of dementia characterised by advanced symptoms in affected individuals' later years. Significant intellectual deficiencies, memory loss, and other cognitive impairments characterise the lives of Alzheimer's patients. Diagnosing Alzheimer's disease may be difficult and time-consuming due to the multitude of mental and physical tests neurologists often use. MCI, or mild cognitive impairment, is a kind of dementia that occurs in the early stages of Alzheimer's disease. The last stage of MCI is called late-MCI, and it is sometimes mistaken for the first stages of Alzheimer's disease (EAD). Correctly classifying EAD is also crucial for preventing or delaying the start of AD. The most fundamental modification in terms of AD's physical presentation is the degeneration of brain cells. Critical biomarkers related with the illness may be uncovered by careful analysis of brain images. The use of magnetic resonance imaging, commonly referred to as an MRI, is a common diagnostic tool used in the medical imaging area during clinical investigations. A large quantity of MRI data was collected from a number of publicly available sources in order to conduct this investigation. All the photos that were taken have had the "skull stripped" effect added to them. The skull and other non-brain pixels carry very little information, hence this is necessary

The Intelligent Technical Influence in Chat Generative Pre-Trained among Students for Modern Learning Traits

2024-03-26

This paper explores the influence of Chat Generative Pre-Trained, which is an Artificial Intelligence tool which provides text-based responses to user queries, on students' learning motivation. The study involved 500 participants who were students in Medan, Indonesia. The research employed a quantitative approach, using surveys and questionnaires to gather data from the respondents. Previous research instruments were adapted with some modifications, which resulting in 10 items for the dependent and independent variables for each. Hypotheses were tested using linear regression, and classical assumption tests, such as multicollinearity, heteroscedasticity, and normality, were conducted. Descriptive statistics, like mean scores, were utilized to assess the extent of Chat GPT usage among students. The results showed that male students displayed a greater propensity for utilizing Chat GPT when compared to their female counterparts. Interestingly, younger students exhibited a higher degree of engagement with Chat GPT in contrast to their older peers. Additionally, the study uncovered a notable, positive, and statistically significant influence of Chat GPT usage on students' motivation to learn.

Prejudge: A Predictive Analytics System for Crime and Legal Judgments

Aastha Budhiraja, Kamlesh Sharma — 2024-03-26

The recent era has seen a substantial inflow of legal documents in the electronic format. Given the fact that data mining can be employed in the world of textual data to extract relevant knowledge, it is being prominently exploited in the domain of criminology and legal matters. With increasing crime rates day-by-day, it has become essential to readily impart justice to the victims. It takes a considerable amount of time for the lawyers to go through previous judgments for their research. The judicial process can be accelerated by decreasing the time spent on research work. Smart legal systems have enormous potential for providing significant insights to the legal community and the general public through the use of legal data. As a result, these systems can assist in the analysis and mitigation of a variety of societal concerns. By extracting numerous things from legal decisions, such as dates, case numbers, reference cases, person names, and so on, this work takes the first step toward realizing a smart legal system. The major research issues in the area of applying machine learning in jurisprudence are information extraction and analysis of legal texts. This study proposes an Machine Learning based framework to improve the user's query for retrieval of precisely relevant legal judgments in order to overcome these limitations. This work has been carried out in order to act as an aid to the legal advisors and the lawyers in framing arguments to make strong standpoints based on predictions given on their case pertaining to previous judicial outcomes for similar such cases. Logistic regression-based classification enables efficient retrieval and prediction by allowing inferences based on domain knowledge collected during the dataset development. According to empirical results obtained, the proposed methodology generates finer results than other traditional approaches.

Artificial Intelligence, Content Recommendation, Biases, and Consumer Behavior: An Analysis of the Impact of Artificial Intelligence on Consumer Behavior

2024-03-26

Purpose: Information plays very important role in decision making. Awareness of brand, price, discounts, post sales activities like guarantee, warrantee, and maintenance must be advertised to influence the buying behaviour. But what if this information creates a bias. Does any type of bias generated by this information, in the form of advertisement, influence the buying behaviour? The present research is exploring the fact that how artificial intelligence-based advertisement suggestions and content recommendations create certain type of bias and how it affects the buying behaviour. This research is based upon a survey of consumers.

Design/methodology/approach: The methodology emphasised to eliminate the errors in measurement. Respondents were approached twice, in a gap of 30 days for collecting data. They were asked to retake the survey and data in both the attempts have been examined for any major deviation. The average of scores have been consolidated as final data of the research analysis.

Findings: The linear regression equation coefficients for the various model variables. The "B" values are the coefficients for each variable. In model 04 we could predict buying behaviour as BB (y) = 0.589 + .403 Anchoring bias + .284 Conformity Bias + .259 Heuristic Bias+ .233 FOMO.

Originality: Researchers have emphasis on exploring a new set of influencing factors for consumer behaviour rather following the key factors in systematic review of previous works. Thus, the work ensures the originality in research.

Role of Computer Mapping in the Strategies of Water Conservation in Green Buildings as per IGBC guidelines- A Case study

Sushma R, Nuthana N, Lakshmi C — 2024-03-26

Green buildings are sustainable or eco-friendly buildings. They conserve natural resources, reduce reliance on fossil fuels and minimize negative environmental impacts. Green building technologies with the help of computers can improve the quality of life for residents and mitigate the environmental and economic challenges associated with rapid urbanization and resource depletion. Conservation of water is a mandatory requirement of green buildings. The amount of water that is suitable for human consumption is only 0.3%. Many reports predict that by 2030, India's water demand will be double the supply. Groundwater, which constitutes 40% of the country's water supply, is depleting rapidly, with 54% of India's groundwater sources in decline. Different strategies are suggested to reduce the water requirements in buildings and landscaping the areas. The Indian Green Building Council (IGBC) claims that techniques including rain gardens, green roofing, rainwater harvesting, recycling, and the reuse of treated wastewater are employed to lower the portable water demand in green buildings. The study discusses the role of computer mapping in water management. The latest graphics processing and display functions are powerful tools and are used in several fields including water conservation. The study put forth that computer intelligence, automation and other characteristics can effectively improve soil and water conservation. The study emphasises the role of computer applications like mapping in water conservation methods followed in green buildings as per IGBC guidelines.

Retracted

Retracted — 2024-03-26

Retracted

Interpretation of Change in Stress Measurement using Strain Gauge during Stressing of Pre-stressing Cables in a Bridge Span

Partha Pratim Roy, Amitava Sil — 2024-03-26

This paper presents a comprehensive study of the interpretation of changes in stress measurements using strain gauges during stressing of prestressed cables in a bridge span. The use of prestressed concrete in bridge construction has become prevalent owing to its ability to counteract tensile stresses and enhance structural integrity. Strain gauges are commonly employed to monitor the stress levels within the prestressing cables during the stressing process. This paper employs computer simulations to analyze and interpret the changes in stress measurements obtained from strain gauges during cable stressing, contributing to a deeper understanding of the structural behavior and load distribution. The study employs a finite element analysis (FEA) to simulate the cable stressing process, providing insights into the stress redistribution within the bridge span. The findings highlight the significance of accurate stress interpretation in ensuring the safety and efficiency of pre-stressed concrete bridge structures.

Sentiment Analysis using a Multinomial LR-LSTM Model

2024-03-26

Sentiment analysis (SA) refers to a technique utilized to ascertain the emotional state conveyed in information or text. It involves categorizing the text into three classes: positive, negative, or neutral. For instance, when someone says "the aqi of the city is good," they are expressing a positive opinion about the aqi of a specific place, while the statement "the aqi is bad" reflects the opposite. The introduction of social media increased the amount of content on the internet of sentiment data. Users on various social media platforms have been able to offer their opinions on various products, services, etc. These opinions are often expressed on social media in the form of movie reviews, product reviews, user comments, posts, etc. In light of this context, one of the captivating research areas in Natural Language Processing (NLP) is Twitter sentiment analysis. The paper proposes a stacked Multinomial-LR-LSTM model for the classification of tweets into three classes. Tweets are re-annotated using Text Blob. Twitter Sentiment dataset was used for experiments with accuracy of 97%.

Ai-Powered Insights into Diabetes Mellitus: A Comprehensive Systematic Review

Vikas J. Magar, Sachin B. Bhoite, Rajivkumar S. Mente, Tulashiram B. Pisal — 2024-03-26

This comprehensive systematic review delves into the current landscape of artificial intelligence (AI) applications to illuminate the intricate metabolic processes and facets of diabetes mellitus. The primary objective is to thoroughly scrutinize and assess the existing body of studies to uncover potential benefits that AI may offer in identifying diabetes mellitus. This study delves into AI's potential for diabetes management, from early detection and personalized therapy to predictive modeling. It critically assesses both the benefits and drawbacks of AI integration, paving the way for responsible future advancements in this complex field. By uncovering AI's potential in diabetes research and exploring its impact on healthcare, this analysis ignites a transformation in how technology shapes both research and treatment.

Optimizing Diabetes Prediction: LDA Pre-processing & ANN Classification in Healthcare`

Soumya K N, Raja Praveen K N — 2024-03-26

Diabetes mellitus (DM) is a chronic disease that poses significant health risks if not well managed. The current healthcare system is overwhelmed by the impact of DM. Modern machine learning and deep learning methods have a hard time correctly predicting the stages of diabetes and often encounter decreased classification accuracy when dealing with massive datasets. In this work, we provide a novel approach to address these problems by integrating pre-processing with LDA and ANN for classification. By combining the LDA and ANN probability distribution functions by back propagation with initialized weights, our method enhances the accuracy of diabetes categorization. After pre-processing data from the PIMA and NCSU datasets using min-max normalization, bivariate filter-based feature selection is used to identify crucial characteristics. Pearson correlation is used to improve the feature set according to a threshold value, further refining the selected qualities. Our experimental results demonstrate the efficacy of the proposed approach, surpassing even the most cutting-edge methods. By integrating a robust classification model with advanced pre-processing techniques, our strategy produces encouraging outcomes in the accurate prediction of diabetes, which in turn helps to improve healthcare management methods.

New Artificial-Based Automated Quality Risk Prediction Methodology for College Students with Disabilitie’s Entrepreneurial Schemes

Hengyun Shen, Zhiyuan Lv, Siti Nisrin Binti Mohd Anis — 2024-03-26

Evaluating and predicting the risk of entrepreneurial projects among college students with disabilities is a critical endeavor that requires a multifaceted approach. This process involves assessing various factors such as the nature of the business idea, the skills and capabilities of the student, potential market demand, and external environmental factors. the issues surrounding entrepreneurial projects among college students with disabilities require a nuanced understanding of the unique challenges they face. Accessibility barriers, societal stereotypes, limited support networks, and lack of inclusive resources are among the key issues hindering their entrepreneurial endeavors. To foster an inclusive environment, it's essential to implement targeted interventions, provide accessible resources and mentorship, raise awareness, and advocate for policy changes that promote equity and accessibility in entrepreneurship for individuals with disabilities. This paper proposed an Automated Quality Risk Prediction (AQRP). The proposed AQRP model uses the Quality assessment of the project at each stage with the ranking-based classification model. The AQRP estimates the process of ranking at every stage of the project and performs the assessment and evaluation of risk. Factors such as the quality of human features in the project and practical features are examined to estimate the features through the process of ranking. With the AQRP model, the features are ranked and integrated for the extraction and classification of features in the projects. With AQRP model the deep learning model is implemented for the classification of features in the projects. Simulation analysis demonstrated that social factors contribute significantly to the project quality assessment. Through the ranking, it is observed that ranking features comprise a higher feature value of 0.98 than the other features. The classification accuracy is achieved as 99% which is 12% higher than the conventional SVM and Linear Regression Classifiers.

Hidden Feature Weighted Deep Ranking Model (Hfwdr): A Novel Deep Learning Approach to Investigate the Nuanced Aesthetic Value of the Elderly Furniture Design & Cultural Identity

Jing Lu，Musdi bin Hj Shanat — 2024-03-26

Retracted

Retracted — 2024-03-26

Retracted

Comparative Analysis of Different Argumentation Frameworks

Shashi Prabha Anan, Vaishali Singh — 2024-03-26

Argumentation Mining is considered a much harder task than generic information extraction or event mining because argumentation structures can be nested recursively. That is, a complete argumentation structure (claim and premises) might function as the premise of some more general claim, and so on. Recognizing the relationships among components of an argument also requires real-world knowledge, including knowing when one thing is a subtype of another. Both use NLP methods to map unstructured text onto graph-like structures or databases. The resulting information is easier to analyze for a variety of tasks, such as learning about social or political views, advising people about how to weigh the evidence for or against some choice, or helping companies to market products or perform quality assurance. Most of these tasks use hand-built templates that have been specified to fit a particular task or observed style of communication.

Developing a Multimodal Deep Learning System for Comprehensive Nutritional Analysis of Meals for Diabetes Management

Kalivaraprasad B, Prasad M.V.D., Bharathi.H. Reddy — 2024-03-26

The management of diet is a pivotal factor in the maintenance of ideal blood glucose levels in individuals diagnosed with diabetes. Precisely evaluating the nutritional value of meals, encompassing caloric intake, can pose a formidable challenge. The present research suggests the creation and implementation of a multimodal deep learning framework aimed at approximating the nutritional composition of meals through the integration of image and textual information. The proposed system aims to combine convolutional neural networks (CNN) for image analysis with recurrent neural networks (RNN) or transformer models for text analysis. This integration is intended to exploit the complementary nature of visual and textual meal data, resulting in more precise estimates. The system is trained using a significant dataset consisting of images of meals, their corresponding textual descriptions, and related nutritional data. This dataset forms the foundation for the system’s development. The model’s predictive accuracy is evaluated through a rigorous assessment on unseen data, utilizing appropriate regression metrics such as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). In addition, we have created a proof-of-concept software application to showcase the practicality of the model in real-world scenarios. The objective of this application is to simplify the process of nutritional monitoring for individuals who have diabetes. The results of this study have the potential to revolutionize dietary management strategies in the context of diabetes care, as they provide a comprehensive and user-friendly nutritional analysis tool. Prospective areas of research encompass enhancing the precision of the model, expanding its scope of food items, and amalgamating it with other healthcare frameworks to achieve a comprehensive approach towards the management of diabetes. .

Identifying Complex Human Actions with a Hierarchical Feature Reduction and Deep Learning-Based Approach

Lakshmi Alekhya Jandhyam, Ragupathy Rengaswamy, Narayana Satyala — 2024-03-26

Among computer vision's most appealing and useful research areas is automated human activity recognition. In these systems, look and movement patterns in video clips are used to classify human behaviour. Nevertheless, the majority of previous research has either ignored or failed to employ time data to predict action identification in video sequences using standard techniques and classical neural networks. On the other hand, reliable and precise human action recognition requires a significant processing cost. To get over the challenges of the pre-processing stage, in this work, we choose a sample of frames at random from the input sequences. We only take the most noticeable elements from the representative frame rather than the entire set of attributes. We suggest a hierarchical approach in which bone modelling and a deep neural network are used first, followed by background reduction and HOG. For selecting features and historical data retention, a CNN along with LSTM recurrent network combo is taken into consideration; in the end, a SoftMax-KNN classification is employed to detect the human behaviours. The name of our model is represented by the abbreviation HFR-DL, which stands for a hierarchical Features Lowering & Deep Learning-based action detection approach. We utilize the UCF101 dataset, which is popular among action recognition researchers, for benchmarking in order to assess the suggested approach. There are 101 challenging tasks in the wild included in the dataset. When comparing the experimental results with eight cutting-edge methods, significant improvements in speed and accuracy are seen.

Reactive & Multipath Routing with Adaptive Urban Area Vehicular Traffic (AUAVT) in VANET Environment

Akanksha Vyas, Sachin Puntambekar — 2024-03-26

VANET is a special instance of the wireless multi-hop system, which owing to the high vehicle mobility is restricted by rapid changes in the topology. In this work family of AUAVT proposed, road-based information routing protocols that performs well in urban areas as adhoc vehicle networks (VANET). AUAVT protocols take advantage of real time traffic generation and communication to construct internets of vehicle (IoV) network. In the proposed work adaptively reactive and multipath AUAVT routing protocol designed and implemented with comparative analysis done on the basis of QoS parameters which is compared with AODV and OLSR routing protocol using NS-2 simulator. Simulation indicates the proposed AUAVT-_Mulp, AUAVT-_Reac routing protocol gives better performance by 6% and 28% respectively in terms of packet delivery ratio and average throughput with less routing overhead over AODV, OLSR routing protocol.

Empowering Collaborative Programming: The Colab Code Strategy for Consistency and Awareness

2024-03-26

The ColabCode is a ground-breaking solution that empowers geographically dispersed individuals working on the same project to collaborate seamlessly and simultaneously. Leveraging advanced technologies and an intuitive user interface, ColabCode provides a comprehensive platform where participants can collaborate, share code, and communicate effectively in real time.ColabCode addresses the challenges faced by distributed teams, allowing them to work cohesively on shared projects. By providing a unified workspace accessible from anywhere, participants can log in to the platform and instantly join the project in progress. Regardless of their location or time zone, team members can start working together, eliminating delays caused by coordination issues. The heart of ColabCode lies in its diverse programming language support. With built-in capabilities for various programming languages such as Python, Java, JavaScript, C++, and more, team members can seamlessly contribute code in their preferred language. This flexibility fosters inclusivity and encourages diverse skill sets, ensuring each member can work efficiently using their expertise.

Machine Learning for Characterization and Analysis of Microstructure and Spectral Data of Materials

Venkataramaiah Gude, Sujeeth T, K Sree Divya, P. Dileep Kumar Reddy, G. Ramesh — 2024-03-26

In the contemporary world, there is lot of research going on in creating novel nano materials that are essential for many industries including electronic chips and storage devices in cloud to mention few. At the same time, there is emergence of usage of machine learning (ML) for solving problems in different industries such as manufacturing, physics and chemical engineering. ML has potential to solve many real world problems with its ability to learn in either supervised or unsupervised means. It is inferred from the state of the art that that it is essential to use ML methods for analysing imagery of nano materials so as to ascertain facts further towards characterization and analysis of microstructure and spectral data of materials. Towards this end, in this paper, we proposed a ML based methodology for STEM image analysis and spectral data analysis from STEM image of a nano material. We proposed an algorithm named Machine Learning for STEM Image Analysis (ML-SIA) for analysing STEM image of a nano material. We proposed another algorithm named Machine Learning for STEM Image Spectral Data Analysis (ML-SISDA) for analysing spectral data of STEM image of a nano material. We developed a prototype ML application to implement the algorithms and evaluate the proposed methodology. Experimental results revealed that the ML based approaches are useful for characterization of nano materials. Thus this research helps in taking this forward by triggering further work in the area of material analysis with artificial intelligence.

A Deep Learning Model for Detecting Bullying Comments on Online Social Media

Renetha J B, Bhagya J, Deepthi P S — 2024-03-26

Youth dominate the online world today and the vast majority access social networks. Around the world, cyberbullying is rampant on social media sites and it has become a serious issue for people of all age groups. The bullying content detection by analyzing textual data in social media dataset is one of the most important parts of this work. The use of Deep Learning in Natural Language Processing has become very prevalent for handling the problem of cyberbullying. A large real-world Twitter dataset is collected for cyberbullying analysis. This work aims to analyze cyberbullying across the social media platform using a deep learning model Long Short-Term Memory Recurrent Neural Network or LSTM RNN and to evaluate its performance. The cyberbullying analysis on Twitter dataset using LSTM RNN gives an accuracy of 86%.

Comparison of Machine Learning Models for Effective Software Fault Detection

Shikha Gautam, Ajay Khunteta, Debolina Ghosh — 2024-03-26

Software defect prediction is a field in software engineering that aims to identify and anticipate defects or bugs in software systems before they occur. The goal is to develop techniques and models that can help software development teams prioritize their testing efforts and allocate resources more effectively. For this purpose, various Machine learning techniques used and these algorithms can utilize various features, such as code metrics, historical defect data, and developer information, to build predictive models. This paper aims to develop a model for software defect prediction using various ML algorithms. Experiments were conducted using the proposed model on KC2 dataset from the NASA PROMISE repository. The Decision tree algorithm achieved 73.28%, Naïve Bayes 83.97%, KNN 80.15%, Support vector Machine 82.44% and Random Forest 80.92% for KC2 dataset. The results demonstrated that the different model succeeded in effectively predicting the defects in PROMISE datasets KC2.

Approaches to handle Data Imbalance Problem in Predictive Machine Learning Models: A Comprehensive Review

Govind M. Poddar, Rajendra V. Patill, Satish Kumar N — 2024-03-26

The business organizations ability to grow and flourish mostly relies on how successfully it understands and utilizes the data it has collected; data has become more vital in today's society. Every company or organization at the present time accumulates massive volumes of data across a range of areas, such as finance, trade, business, and healthcare. Medical data may be provided by clinics, doctors, healthcare providers, and insurance establishments. Upon locating the necessary medical datasets, the next phases would be to investigate and utilize appropriate modeling algorithms to mine substantial information for probable prediction. Biased data is significant challenge in machine learning where the distribution of data elements in a dataset is uneven, with one class considerably outnumbering the others. This occurrence leads to biased models and reduced performance that affects quality and reliability of machine learning algorithms. This paper presents detailed review on reasons for imbalanced data, its impact, algorithmic procedures to handle unevenly distributed data. We explore various techniques, algorithms to address problem, advantages, demerits and evaluation metrics to assess performance of procedures for handling imbalanced datasets.

Adversarial Attacks and Defenses in Deep Learning Models

2024-03-26

This paper investigates the complex interactions that lead to adversarial weaknesses in deep learning systems. This analyses various adversarial attack strategies, including FGSM and PGD, to evaluate how well they may undermine model fidelity. These results highlight the ongoing cat-and-mouse game between deep-learning security attackers and defenders. Although much progress has been made in increasing model resilience, the lack of a globally defined strategy highlights the necessity for a diversified security policy. This study shows the need for continual innovation and the persistent difficulty of protecting deep learning models against hostile threats

Anomaly Detection in Time Series Data Using Deep Learning

2024-03-26

This paper investigates anomaly identification in historical data using advanced deep learning algorithms. Traditional methods of statistics, while useful, frequently fail to capture complex temporal connections. Our research thoroughly assesses the success rate of various deep learning structures for this job, including neural networks with RNNs, LSTMs, and CNNs. To refine the data, optimized preprocessing approaches such as normalization, in addition detrending, as well as the engineering of features is used. The models' adaptability and robustness are demonstrated through empirical validation in a variety of areas, including banking, health care, especially industrial processes. The study emphasizes scalability and processing efficiency to ensure practicality in real-world applications. Furthermore, interpretability methods provide perspectives into the machines' decision-making processes. The results reveal that deep learning models outperform conventional methods, paving the path for improved anomaly identification in time series information. Future study recommendations involve looking into hybrid structures, improving model comprehension, and researching real-time anomaly identification approaches. This work advances anomaly detection algorithms, which could have applications ranging from espionage to maintenance forecasting. The optimized framework offered here has the potential to improve system reliability as well as safety across a wide range of sectors.

Blockchain Technology for Secure and Trustworthy Decentralized Applications

2024-03-26

This study explores the complex world of blockchain-based decentralized applications (DApps), concentrating on security and trust mechanisms. The study detects and classifies security flaws in DApps, which include flaws in smart contracts, difficulties with consensus, including dangers of data manipulation. By contrasting reputation-based systems with token-based incentives, it investigates trust mechanisms and clarifies their effects on user behaviour. The study emphasizes the crucial part that blockchain integration plays in boosting DApp security by employing its built-in immutability, decentralization, and cryptographic characteristics. The benefits of blockchain are supported by empirical data and vivid case examples. The paper ends with advice for DApp creators that emphasizes secure development methods, thorough audits, as well as user education while also indicating potential directions for future research.

Design and Develop A Secure Energy Efficient Data Transmission Technique for Wireless Sensor Networks

Avneesh Gour, Nishant Kumar Pathak — 2024-03-26

Wireless Sensor Networks (WSNs) play a vital role in various applications such as environmental monitoring, healthcare, and smart cities. However, energy consumption is a critical concern in WSNs due to the limited power supply of sensor nodes. This research paper proposes a novel approach to address this challenge by designing and developing a secure energy-efficient data transmission technique for WSNs. The proposed technique aims to minimize energy consumption while ensuring data confidentiality, integrity, and authenticity. By leveraging cryptographic algorithms, optimization strategies, and intelligent routing protocols, the proposed technique enhances the security and efficiency of data transmission in WSNs. Experimental results demonstrate the effectiveness and feasibility of the proposed approach in improving the overall performance of WSNs in terms of energy consumption, communication overhead, and security.

A Comprehensive Survey of Multiple Object Tracking Techniques

Hardik Jaiswal, Aditya Gambhir, Laxmi Bewoor, Nagaraju Bogiri — 2024-03-26

Multiple Object Tracking (MOT) is crucial in computer vision and surveillance, especially for automating traffic control in challenging traffic environments. This review surveys advancements in object detection, tracking algorithms, lane departure warnings, and semantic segmentation, with a specific focus on traffic law enforcement. It covers issues like wrong-way, clearway, and one-way traffic violations, as well as challenges including occlusion and splits. Various methods, such as background subtraction and deep learning, are explored.The review stresses the significance of analyzing recent literature for researchers to bridge gaps, overcome limitations, and create new algorithms. It also touches on hardware, datasets, metrics, and research directions. Future MOT research aims to develop efficient algorithms for dynamic tracking, improve detection accuracy, and reduce real-time processing. The survey's proposed methods offer valuable references for tracking multiple objects in frame sequences.

Algorithmic Modeling for Predicting Carbon Emissions in an Individual Vehicles: A Machine Learning and Deep Learning Approach

Rashmi B. Kale, Nuzhat Faiz Shaikh — 2024-03-26

This study proposes an algorithmic model aimed to accurately predicting carbon emissions from individual vehicles by leveraging machine learning and deep learning techniques. Concerns regarding environmental sustainability and climate change have intensified the need for precise assessments of carbon footprints, particularly in the transportation sector. Traditional methods often lack the adaptability and scalability required to handle the complexity of emission prediction tasks. In contrast, machine learning and deep learning offers promising avenues for developing robust models capable of learning from vast datasets and capturing intricate patterns in vehicle emissions. The purpose of research is to address the breach by designing a deep learning algorithmic framework that integrates machine learning algorithms to analyze real time datasets with vehicle attributes, driving patterns, and fuel characteristics to predict carbon emissions. The proposed approach holds potential for enhancing our understanding of vehicle emissions dynamics and facilitating the development of targeted interventions to mitigate environmental impacts.

Pharmaceutical Sales Forecasting with Machine Learning: A Strategic Management Tool for Decision-Making

2024-03-26

This investigation explores the adequacy of machine learning strategies for pharmaceutical deal estimating, displaying a comparative investigation of four calculations: Random Forest, Gradient Boosting, Long Short-Term Memory (LSTM), and AutoRegressive Integrated Moving Average (ARIMA). Real-world pharmaceutical deals information was utilized to assess the prescient execution of these calculations utilizing measurements such as Cruel Absolute Error (MAE), Mean Squared Error (MSE), and Root Cruel Squared Error (RMSE). The results demonstrate that LSTM beats the other calculations, accomplishing the most reduced MAE of 900, MSE of 13000, and RMSE of 113.96. Moreover, the research gives a comprehensive survey of later progressions in prescient analytics and machine learning over different divisions, counting healthcare, supply chain administration, back, and natural supportability. The discoveries emphasize the transformative potential of progressed analytics in driving key decision-making, optimizing asset assignment, and relieving dangers in pharmaceutical deals. Moving forward, the integration of machine learning-driven determining models into organizational procedures will proceed to revolutionize the pharmaceutical industry and clear the way for maintainable development and advancement.

Real-time Anomaly Detection in Big Data Streams: Machine Learning Approaches and Performance Evaluation

2024-03-26

The paper focuses on real-time anomaly detection in big data streams, discussing the machine learning techniques as well as performance evaluation. Different outlier detection methods, including Isolation Forest, Local Outlier Factor, Support Vector Machine and Elliptic Envelope, are considered and analyzed relying on a dataset of financial transactions. The analysis is made up of the visualization of data distributions, correlation matrices and metrics like accuracy and classification reports. The findings reveal varying performance levels among the methods pointing out the significance of choosing appropriate techniques for effective anomaly detection in dynamic data environments. This paper adds to the knowledge of outlier detection in big data streams and provides valuable insights for future studies.

Practical Implementation of Blockchain Technology in Pharmaceutical Supply Chain Management

Rishi JP, Ramdas Bhat, Prateek Srivastava, M. Sundar Raj — 2024-03-26

This research examines the viable usage of blockchain innovation in pharmaceutical supply chain administration. Through an arrangement of tests and comparative examinations, the ponder assesses the effect of blockchain on key execution pointers such as traceability, straightforwardness, security, and proficiency. Results show noteworthy changes in traceability, with blockchain empowering real-time following of pharmaceutical items from fabricating to conveyance. Straightforwardness is improved through the utilisation of decentralized records, giving partners perceivability into supply chain operations. Security measures such as cryptographic hashing and computerized marks guarantee information keenness and ensure against extortion and unauthorized get to. Besides, the mechanization of forms by means of shrewd contracts leads to expanded proficiency in exchange preparation and compliance confirmation. The discoveries of this investigation highlight the transformative potential of blockchain innovation in revolutionizing pharmaceutical supply chain administration. Moving forward, it is basic for partners to address challenges such as adaptability and administrative compliance to completely realize the benefits of blockchain in guaranteeing the secure and effective conveyance of pharmaceutical items.

Integrating Artificial Intelligence and Data Analytics for Supply Chain Optimization in the Pharmaceutical Industry

P Kiran Kumar Reddy, Atish Mane, Atowar ul Islam, Reecha Singh, Fahmida Khatoon — 2024-03-26

This inquire about examines the integration of Artificial Intelligence (AI) and information analytics to optimize supply chain forms within the pharmaceutical industry. Through tests and writing audits, the ponder investigates the adequacy of AI calculations counting Linear Regression, Random Forest Regression, K-Means Clustering, and Deep Learning Neural Systems over request estimating, stock optimization, generation planning, and coordination optimization. Results appear that Random Forest Relapse beats Direct Relapse in request determining with RMSE of 80.20, MAE of 60.75, R² of 0.90, and MAPE of 6.50%. K-Means Clustering recognizes five clusters for stock optimization. Profound Learning Neural Systems accomplish RMSE of 75.10, MAE of 55.30, R² of 0.92, and MAPE of 5.80% for generation planning. In coordination’s optimization, Genetic Algorithm accomplishes a add up to fetched of $150,000 and conveyance time of 5 days compared to Mimicked Strengthening with $160,000 and 6 days. The research contributes to understanding the part of AI and information analytics in improving supply chain effectiveness, decreasing costs, and guaranteeing maintainability within the pharmaceutical segment.

"Implementing AI-Driven Personalized Medicine in Clinical Practice: Challenges and Practical Solutions"

Sweety Bakyarani E, Virender Kumar Dahiya, Yogita Bhise, Subramanian Selvakumar — 2024-03-26

This investigation explores the usage of AI-driven personalized pharmaceuticals in clinical hone, tending to challenges and proposing arrangements. Leveraging calculations counting Bolster Vector Machines, Random Forest, Neural Networks, and Bayesian Systems, it assesses their viability in optimizing treatment methodologies and improving quiet results. Experimentation on a differing dataset uncovers Neural Networks as the foremost compelling, accomplishing 90% precision, 88% affectability, and 92% specificity. Comparison with existing writing highlights the transformative potential of AI over healthcare spaces, emphasizing morals, security, infection administration, and restorative instruction. The consideration underscores the significance of tending to moral, administrative, and socio-cultural variables for belief and acknowledgement of AI in medication. Future investigate ought to approve these come about on bigger datasets and address real-world usage challenges. By grasping AI as a complementary apparatus in clinical workflows, healthcare suppliers can upgrade care conveyance, eventually progressing personalized medication and making strides health results.

Cloud-Enabled Social Network Mining for Advanced Recommendation Systems: An Integrated Data Mining and Social Network Analysis Approach

Jaishree Jain, Santosh Kumar Upadhyay, Sharvan Kumar, Neerja Arora — 2024-03-26

Recommendation systems play a crucial role in assisting users in discovering relevant and personalized content in social networks. For recommendation system data mining has a profound impact on several domains, including databases, artificial intelligence, machine learning, and social networks. It plays a crucial role in driving significant research advancements in the field. In today's fast-paced world, where data is rapidly expanding and information retrieval poses complex challenges, users increasingly demand valuable insights from their vast datasets. Social networks have emerged as a fascinating domain that has made substantial contributions to data mining research, ushering in a new era of possibilities. To determine the intrusion index based on the source address of the network security alarm, a simulation test is run. The findings demonstrate that this strategy can successfully implement cloud network security situation awareness as the related window attack index drops as soon as the security event is cancelled. You can accurately detect changes in network security circumstances using the suggested technique

A Comprehensive System for Sustainable Tree Plantation and Growth Monitoring using Blockchain, AI, and IoT

2024-03-26

There are several environmental challenges faced by the world today, with deforestation and climate change being major threats to the environment and its sustainability. NGOs and Government bodies play a crucial role in addressing these issues by organizing and conducting tree plantation drives. However, a lack of transparency, mismanagement of funds, and inefficient tracking systems, have hindered the effectiveness of these efforts. Many problems occur after tree plantation as there is no record being held to track the growth of trees, funds transparency is not available, no overall analysis is provided for deciding which tree species should be planted in a particular area to achieve maximum sustainability and also to improve the chances of growth of trees. Only planting trees in large numbers won’t help to solve this problem, a proper system is needed which can record time to time data regarding each and every tree through which we can help in the survival of all the trees and increase their lifespan. This will also help us in avoiding the drying and death of trees. The solution that we propose in this paper, to address the existing drawbacks is to create a web3 based platform to ensure the transparency of transferred funds and tree plantation by NGOs and government bodies, along with which we will implement a feature of tracking the status of planted trees using volunteers and IOT device in areas that aren’t easily accessible by volunteers

Adaptive Dragonfly Optimization (Ado) Feature Selection Model and Distributed Bayesian Matrix Decomposition for Big Data Analytics

M.Vijetha, G.Maria Priscilla — 2024-03-26

Matrix decompositions are fundamental methods for extracting knowledge from large data sets produced by contemporary applications. Processing extremely large amounts of data using single machines are still inefficient or impractical. Distributed matrix decompositions are necessary and practical tools for big data analytics where high dimensionalities and complexities of large datasets hinder the data mining processes. Current approaches consume more execution time making it imperative to reduce dataset feature counts in processing. This work presents a novel wrapper feature selection method utilising Adaptive Dragonfly Optimisation (ADO) algorithm for making the search space more appropriate for feature selections. ADO was used to transform continuous vector search spaces into their binary representations. Distributed Bayesian Matrix Decomposition (DBMD) model is presented for clustering and mining voluminous data. This work specifically uses, 1) accelerated gradient descent, 2) alternate direction method of multipliers (ADMM), and 3) statistical inferences to model distributed computing. These algorithms' theoretical convergence behaviours are examined where tests reveal that the suggested algorithms perform better or on par with two common distributed approaches. The methods also scale up effectively to large data sets. Clustering performances are assessed using the metrics of precision, recall, F-measure, and Rand Index (RI), which are better suited for imbalanced classes.

Predicting Landslides through Satellite Imagery Analysis and Machine Learning

Anup Kadu, Raj Mishra, Vishal Shirsath — 2024-03-22

The effects of climate change on landslides become more apparent, this work presents a novel method of landslide prediction by combining cutting-edge machine learning algorithms with Google Earth satellite images. Using digital image processing and Geographic Information System (GIS) techniques, the proposed method extracts important parameters, like elevation and slope, from high-resolution satellite data. Landslides are becoming a critical threat due to their increasing frequency, necessitating accurate prediction and early warning systems. Then, an intricate digital elevation model (DEM) is created and utilized as an input for more complex machine learning models, such as CNN and polygonal neural networks. Precise prediction of probable landslide events across large, susceptible areas is made possible by this novel combination. Landslides may have a major negative impact on human life and the economy, but the integrated method greatly improves the accuracy of early detection. The results demonstrate the efficacy of this innovative approach in delivering precise and timely forecasts, signifying a significant advancement in the evaluation of geotechnical hazards and proactive risk control for expansive, high-risk regions. In order to meet the urgent need for proactive mitigation in the face of climate-induced risks, this research presents a strong foundation for comprehensive landslide prediction.

Machine Learning for Precision Agriculture: Predictive Analysis of Crop Growth Frequencies

Niketa Kadam, Raj Mishra, Vishal Shirsath — 2024-03-22

By combining cutting-edge machine learning techniques with the examination of plant development patterns, presented research employs an innovative dual methodology for precisely calculating green areas, i.e., plant growth. One unique aspect of the research is the correlation between plant development and specific musical frequencies, which span from 1 to 10 kHz and include pop, classical, and Normal. By utilizing support vector machines (SVM) and artificial neural networks (ANN), the dual method improves our comprehension of the dynamics of plant development. Interestingly, the study shows that SVM performs better than ANN, offering more accuracy in predicting green areas. This sophisticated approach shows how fusing state-of-the-art neural networks with conventional machine learning may revolutionize the field and change the course of precision agriculture. The study highlights the complementary nature of contemporary and traditional methods, demonstrating their effectiveness in providing a thorough understanding of plant development and a productive assessment of green areas. SVM's astounding accuracy levels up to 92.10% highlight the significance of this technology in the advancement of precision farming practices. The stability and applicability of proposed strategy are emphasized, especially in light of accurate and successful agricultural management techniques. Three plant species were watched over the course of three months for this study, giving the results a strong real-world component.

Risk Inference-Based Privacy Preservation Model for Cyber-Physical Systems

Manas Kumar Yogi, A. S. N. Chakravarthy — 2024-03-22

The Risk Inference-Based Privacy Preservation Model for Cyber-Physical Systems (CPS) addresses the critical need for safeguarding privacy in the evolving landscape of interconnected physical and digital environments. This model, aptly named RIP2 (Risk Inference for Privacy-Preserving CPS), integrates advanced risk assessment techniques with robust privacy-preserving mechanisms to create a dynamic and adaptive framework. The model begins with a comprehensive risk assessment module that identifies potential threats, values privacy-sensitive assets, and assesses vulnerabilities within the CPS architecture. A privacy risk inference engine dynamically analyses contextual data, user behavior, and continuously evolving risk factors to assess the current privacy risk level. Privacy-preserving mechanisms, including differential privacy, encryption, and anonymization, are adaptively applied based on the inferred risk level, ensuring a tailored and effective approach to privacy preservation. Users are empowered to define their privacy preferences, and the model incorporates dynamic privacy policies that automatically adjust based on the risk assessment. Furthermore, the model incorporates incident response and continuous learning mechanisms to respond promptly to privacy incidents and improve the overall resilience of the system. The RIP2 Model aims to strike a balance between the seamless functionality of CPS and the paramount importance of preserving individual privacy in an interconnected and data-driven world.

Investigation of SAR Reduction and Bending Effect Using a Flexible Antenna with EBG Structure for 2.45 GHz Wearable Applications

Sonal Jatkar, Nilesh Kasat — 2024-03-22

This paper presents, an analysis of Co-Planar Waveguide fed textile antennas operating at 2.45 GHz frequency with an Electromagnetic Band-Gap array for WBAN applications. Denim substrate, characterized by its dielectric constant of 1.6 and thickness of 1mm which is stacked together to get height of 3mm. Simulation using HFSS provides a return loss of -52 dB at 2.45 GHz. Size of antenna after incorporating 2*2 array in EBG structure is 61*61*3 mm³. It results in increase gain from 2.36 dB to 4 dB without EBG and with EBG respectively. Stochastic frameworks employing polynomial chaos principles expansions is used to model antenna dimensions. The study aimed to investigate how bending affects antenna performance, employing two bending scenarios: one in the H plane and the other in the E plane. Also, for WBAN applications, SAR value is simulated on human body phantom using HFSS simulator and is obtained to 41.04 W/Kg without EBG and 0.522 W/Kg with EBG.

AI-Enabled Customer Relationship Management: Personalization, Segmentation, and Customer Retention Strategies

2024-03-22

Artificial intelligence (AI) and machine learning are transforming customer relationship management (CRM) strategies. This paper provides an extensive review of how AI-enabled capabilities like predictive analytics, personalization engines, and customer segmentation are enabling more tailored, relevant experiences that strengthen customer relationships and loyalty over time. Current CRM systems generate massive datasets on customer interactions and behaviors, which feed AI algorithms to uncover hidden insights around individual preferences, likely future behaviors, and optimal cross-sell recommendations unique to each customer. We analyze key AI methodologies powering next-generation CRM including reinforcement learning, neural networks, natural language processing, and computer vision. The paper discusses sample use cases and real-world examples of AI-driven CRM initiatives from leading companies that focus on personalization, predictive churn models, next-best action recommendations, and automated customer service agents. We also examine emerging technologies on the horizon such as affective computing, virtual reality, and the metaverse that present new opportunities to understand customers and meet their needs in highly tailored, emotionally intelligent ways. The paper concludes with an analysis of critical considerations as firms implement AI-enabled CRM including data privacy, transparent AI, and avoiding algorithmic bias. With responsible implementation, AI stands poised to revolutionize CRM with previously impossible levels of personal relevance at scale, ultimately growing customer lifetime value.

Robotics and Cobotics: A Comprehensive Review of Technological Advancements, Applications, and Collaborative Robotics in Industry

2024-03-22

Collaborative robotics, or cobots, are transforming human-robot interaction in industrial environments. This paper provides a comprehensive review of the technological advancements, applications, and collaborative aspects of robotics across various industry verticals. Advanced hardware and software innovations are enabling robots to work safely alongside humans, enhancing productivity and quality while also taking over undesirable or dangerous tasks. Cobots are being rapidly deployed for assembly, pick and place, inspection, machine tending and other precision handling operations. Implementation challenges exist, but continued improvements in sensing and intelligence capabilities are increasing robot flexibility and ease of integration in human-centric work cells. With appropriate configuration, deployment strategies and worker training, collaborative robots can improve manufacturing and production performance. This paper examines the rise of collaborative industrial robots and analyzes the outlook for this technology over the next five years.

Assessment of Conflict Flows in Software-Defined Networks using a Novel Nature-Inspired Optimization-Tuned Kernelized SVM

2024-03-26

The centralizing management and flexibly customizing network resources, Software-Defined Networks (SDN) completely transform network administration. As networks become more intricate, the likelihood of conflicts arising data flows increases, potentially leading to a decline in overall performance and the emergence of security vulnerabilities. This paper presents a tree-seed optimization-tuned kernelized support vector machine (TSO-KSVM) for the assessment of conflict flows in SDN environments. Initially, we gather data samples of SDN in conflict flows to analyze the performance of the proposed method. Applying the min-max scaling method to preprocess the raw data samples and linear discriminant analysis (LDA) is carried out to reduce the dimension. In the proposed framework, TSO is applied to enhance the assessment in the KSVM model. The proposed method is implemented in the Python tool. The proposed method's performance is analyzed in terms of various metrics compared with other methods. From the experimented results, we conclude that the proposed method attains the greatest accuracy rate of other methods in assessing conflict flows in SDN networks.

Novel Resource Allocation Approach for Fog Computing-Driven IoT Systems

2024-03-26

Fog computing (FC) has the potential to lower latency and boost speed. Internet of Things (IoT) networks have difficulties allocating resources efficiently. The approaches used are flexible, scalable, or optimized. To maximize performance indicators, new approaches that utilize real-time information, workload sequences, device accessibility and network circumstances are required. We investigate the allocation of resources and task scheduling for numerous devices in IoT systems in this research. IoT devices must properly choose which data to offload to FC nodes (FCNs) as they acquire enormous amounts of data. To tackle the problem of supporting multiple device connections and facilitating fast data transfers with constrained resources, we suggest executing non-orthogonal multiple access (NOMA). Several devices can simultaneously send data spanning time, frequency and coding domains to an identical FCN because of NOMA. Together, we optimize power transmission and resource assignment for IoT devices, meeting QoS requirements and reducing network energy usage. In this research, a unique boosted atom search optimization (BASO) method is presented to tackle it because it is an NP-hard issue. According to the simulation results, the suggested strategy outperforms in terms of greatest throughput, minimum latency and optimal energy use.

Revolutionizing Image Encryption: Data Hiding Model Based on Optimized Neural Network

2024-03-26

Information security researchers are currently focusing on Revolutionizing Image Encryption in data hiding for safe digital data movement. Preserving information that has been hidden in a system is one of the core concepts of data hiding. For image encryption, data hiding is the process of embedding private data into images to ensure it is hidden from perceptions by other people. In this research, the hidden data is retrieved by our proposed Multi-rate Salient Gated Recurrent Neural Network (MSG-RNN) and it employs a dependent classification method to recover images from encrypted images. We gathered a data of various kinds of image data. Following the encryption of the original image, we established the Elliptic Curve Cryptography (ECC) method and created an innovative image encryption technique to improve security by data hiding. We calculated our proposed method's bit error rates. The comparison evaluation is performed with various methods to estimate proposed technique. Using the MSG-RNN approach on a number of images produced superior outcomes and they include the results for boats (0.0714), peppers (0.0712), baboons (0.0716) and airplanes (0.0719). The experimental outcomes offered that the proposed MSG-RNN technique performed better than other existing methods in data hiding process to enhance image encryption.

Developing an Innovative Machine Learning Integrated Cloud Monitoring System for Cloud-Based Services

2024-03-26

A cloud monitoring system is an integrated solution for monitoring as well as managing the performance, availability along with the security of cloud-based infrastructure and services. It employs a variety of tools and methods to gather, analyze and present data on resource use, application activity and user interactions. By monitoring critical parameters in real-time, it allows proactive problem identification and resolution, resource allocation optimization, as well as adherence to service-level agreements. In this research, we developed an innovative machine learning (ML) integrated cloud monitoring system named Sea Horse fine-tuned Extreme Gradient Boosting (SH-XGBoost). Initially, we collected a dataset that includes various types of cloud environment scenarios to train our proposed approach. We utilized the Robust Scaling (RS) algorithm to pre-process the gathered raw data. We employed the Sea Horse Optimization algorithm to enhance the primary characteristics of the proposed XG-Boost algorithm. The suggested approach is implemented in Python software. The finding evaluation phase is performed with multiple metrics such as,F1-score (98.49%), Recall (98.38%), Precision (98.53%) and Accuracy (98.43%) to assess the proposed SH-XGBoost approach with other conventional approaches. The experimental findings illustrate that the proposed SH-XGBoost approach performed better than other existing approaches for novel cloud monitoring systems.

Automated IoT-Based Monitoring and Control for Hydroponic System

Vaira Muthu K., Krishnakumar A. — 2024-03-26

Agriculture, the world of farming is an essential area where the people around are focusing to develop to enhance more yields in minimum cost and other requirements. The new emerging technique named hydroponics focuses in developing a greenhouse that involves developing crops using water-based nutrients without soil. This proposed implementation presents an intelligent design that comprised with low-cost and automatic monitoring control through the support of IoT (Internet of Things) for hydroponics greenhouse. This implementation includes some sensors to monitor and controls pumping of water, a quality of water, monitor the temperature and humidity of the crops. The master node controls the water flow and aggregates the data, which is received from the member nodes. Member nodes monitor the temperature and humidity and forwards the data to the master node for necessary actions. A Fuzzy inference model is proposed to determine the flow of water and nutrients. The proposed model outperforms than the existing model in low cost, better energy efficiency and throughput.

Enhancing Maximum Power Point Tracking through Ensemble Techniques

Hayder Husam Mahmood, Zaid Hamodat — 2024-03-26

Maximum Power Point Tracking (MPPT) plays a pivotal role in photovoltaic (PV) solar systems, streamlining the harnessing of available power and bolstering energy conversion efficiency. Its significance lies in its alignment with the global push to heighten the efficacy of renewable energy sources. This article unfolds a meticulous examination of the predictive modeling specific to solar energy. The investigation spans various machine learning models such as Linear Regression (LR), Support Vector Regression (SVR), XGBoost Regressor, and Ensemble Learning (EL), each dissected to reveal the intricacies involved in solar energy system modeling. The research, conducted across two unique datasets—Solar Power Generation and Solar Radiation Prediction, employed rigorous statistical evaluation to uncover the distinctions in accuracy, unity, and efficacy among the models. A standout finding was the Ensemble Learning model's superior performance, notably through applying techniques like Bagging Regressor. This approach transcended the individual models in both datasets by ingeniously amalgamating the predictions of various underlying models, leading to enhanced predictive precision. This article's insights contribute considerably to the domain of solar energy modeling, elevating Ensemble Learning as a powerful instrument for refining prediction accuracy. Furthermore, the juxtaposition of various modeling methodologies unveils valuable insights into their respective trade-offs, enriching the foundation for future exploration and real-world implementations within the renewable energy landscape. In setting a novel standard in solar energy forecasting, this study also resonates with the broader objectives of sustainable energy governance and ecological preservation.

Optimizing Microgrid Performance: A Data-Driven Approach with IoT Integration

Raafat K. Oubida — 2024-03-26

The essential goal of this exploration is to work on the efficiency of microgrids by using a data-driven system that is expanded by the joining of Internet of Things (IoT) technology integration. Microgrids, which are decentralized energy systems, are a fundamental part during the time spent upgrading energy resilience, bringing down fossil fuel byproducts, and further developing admittance to power. By and by, there are serious issues associated with expanding the efficiency and reliability of microgrid tasks. These difficulties are brought about by the multifaceted cooperations that happen between the a large number and the dynamic outer impacts. Inside the extent of this examination, we offer a remarkable system that utilizes data-driven procedures to evaluate, reproduce, and improve the performance of microgrids. We plan predictive models to conjecture energy utilization, further develop energy generation and conveyance, and lift system constancy by using real-time data gathered from Internet of Things (IoT)- empowered sensors and devices implanted inside the design of the microgrid. The consolidation of Internet of Things technology makes it conceivable to consistently screen and work the resources of a microgrid, which thus pursues it more straightforward to settle on proactive decisions and execute adaptive management strategies. Through the use of case studies and simulated tests, we illustrate the efficacy of our method in terms of promoting energy sustainability, lowering operational costs, and improving the performance of microgrids. The findings of this study provide a significant contribution to the development of intelligent energy systems and lay the groundwork for future advances in the optimization and administration of microgrid electricity networks.

A Novel Medical Decision Support System Using Swarm Intelligence Based Bayesian Learning Algorithm

2024-03-26

The use of Machine Learning (ML) methods may be beneficial at the clinical and diagnostic levels of medical decision-making. A foundation for ML is provided by feature selection algorithms. In a medical setting, feature selection may be used to rapidly and efficiently identify the health-related qualities that are most distinctive from the original feature collection. The two primary objectives of feature selection algorithms are to determine the properties of data classes that are most relevant and to enhance classification performance. In addition to assisting lower the general measurement of the dataset, feature selection also aids in determining which features are most important. Therefore, we provide a unique ML-based approach in this study. The dataset is first gathered and prepared using the min-max normalization approach. The features are selected using principal component analysis (PCA). Using a novel swarm-optimized Bayesian learning approach (SOBLA), accuracy is used to evaluate the effectiveness of various feature subsets. Experimental results show that the performance of the proposed method performs better when compared to conventional methods. The outcomes of this study suggest interventions with the potential to enhance the quality of healthcare decision-making about certain healthcare procedures.

Tuna-Osprey Optimization for Energy Efficient Cluster-based Routing: Modified Deep Learning for Node's Energy Prediction

Gopala T, Raviram V — 2024-03-26

The main consideration of WSN design is maximization of the network lifetime. It is proven that the effective balancing of network energy consumption along with the maximization of network lifetime can be performed by clustering and routing approaches. Accordingly, a new Tuna Osprey Optimization algorithm for energy-efficient cluster-based routing has been developed in this work. This approach includes 2 working phases: clustering and routing process. Initially, a modified DL model named M-LSTM is proposed for predicting the node’s energy. Subsequently clustering process is carried out by the TOO algorithm, which considers the energy, link lifetime, distance, trust, and delay as constraints for the selection of optimal CH. Finally, with the same TOO algorithm, the routing process is conducted, which considers the link quality as the constraint to provide optimal routing. Results proved that the proposed TOO for energy-efficient cluster-based routing can reduce energy utilization while attaining maximum network lifetime.

Enhancing Cybersecurity with ML: A Multi-Algorithm Approach to Anomaly-Based Intrusion Detection

Indira P. Joshi, Vijaya K. Shandilya — 2024-03-26

In the era of escalating cyber threats, the significance of robust intrusion detection systems (IDS) cannot be overstated. Traditional methods often struggle to keep pace with the evolving tactics of malicious actors. This paper presents a novel approach to enhancing cybersecurity through the integration of machine learning (ML) techniques within anomaly-based intrusion detection systems. Specifically, we propose a multi-algorithm framework that leverages the complementary strengths of various ML models to effectively identify diverse cyber threats. Our approach aims to address the limitations of single-algorithm systems by combining the capabilities of multiple classifiers. We demonstrate the efficacy of our methodology through extensive experimentation on real-world network traffic scenarios. Results indicate that our multi-algorithm approach outperforms traditional single-algorithm solutions in terms of detection accuracy, false positive rates, and scalability. Furthermore, we discuss the practical implications of our framework in bolstering cybersecurity defenses across diverse organizational contexts. Overall, this research contributes to the advancement of anomaly-based intrusion detection systems by offering a robust and adaptable ML-driven solution capable of effectively combating emerging cyber threats.

Enhancing Network Security through Machine Learning Based Intrusion Detection Systems

2024-03-26

The increasing complexity and sophistication of cyber threats have necessitated the development of robust and intelligent security mechanisms to safeguard network infrastructures. In recent years, machine learning (ML) techniques have emerged as a powerful tool for enhancing network security, particularly in the realm of intrusion detection systems (IDS). This research paper explores the application of machine learning algorithms in the context of IDS to enhance network security. It investigates various ML techniques, their benefits, and challenges, and provides insights into the integration of ML-based IDS in modern network architectures. The study also highlights the potential limitations and future research directions in this evolving field.

Explainable AI for Trustworthy Decision-making in IoT Environments

2024-03-26

Artificial Intelligence (AI) is being widely incorporated into Internet of Things (IoT) contexts, leading to improved automation and efficiency. Given that AI algorithms have a substantial impact on decision-making processes in these intricate ecosystems, it becomes crucial to prioritize their reliability. Explainable AI (XAI) is becoming more important for promoting openness and accountability in decision-making inside the Internet of Things (IoT). Through the provision of explanations that humans can comprehend, explainable artificial intelligence (XAI) improves stakeholders' understanding of the underlying logic and allows for the detection and reduction of any biases or mistakes. This abstract explores the importance of Explainable Artificial Intelligence (XAI) in facilitating reliable decision-making in Internet of Things (IoT) contexts. It highlights the function of XAI in improving transparency, reducing risks, and building trust among stakeholders. This abstract emphasizes the crucial need to include explainability into AI-driven decision-making processes in order to guarantee their dependability and ethical soundness. It does so by thoroughly examining XAI concepts and techniques specifically designed for the problems posed by IoT ecosystems.

Deep Reinforcement Learning for Dynamic Resource Allocation in IoT-enabled Big Data Networks

2024-03-26

In the area of Internet of Things (IoT)-enabled big data networks, the dynamic and diverse character of these settings presents a significant problem in terms of the optimal allocation of resources. Deep Reinforcement Learning (DRL) has emerged as a viable technique to overcome this issue by dynamically adjusting resource allocation algorithms depending on changing network circumstances and demands. This approach has the potential to handle other problems as well. The purpose of this study is to provide a complete assessment and analysis of traditional research efforts that revolve around the use of DRL approaches for dynamic resource allocation in big data networks that are enabled by the Internet of Things (IoT). Furthermore, we emphasize the possible advantages and limits of applying DRL in such complex systems by analyzing the techniques, problems, and successes of previous research that have been conducted in this field. We have identified important research gaps and potential for future investigations via this study. These studies are focused at enhancing the efficacy and scalability of DRL-based resource allocation solutions in big data networks that are enabled by the Internet of Things (IoT).

Machine Learning-based Predictive Analytics for Blockchain-enabled IoT Systems

2024-03-26

In the realm of blockchain-enabled IoT systems, machine learning-based predictive analytics serves as a cornerstone for optimizing operations, enhancing security, and maximizing efficiency. By leveraging the wealth of data generated by IoT devices and immutably recorded on the blockchain, predictive analytics algorithms can discern patterns, detect anomalies, and forecast future events with remarkable accuracy. One key application lies in anomaly detection, where machine learning models scrutinize data to identify aberrant behavior or potential security threats in real-time, thereby fortifying the integrity of the system. Moreover, predictive maintenance emerges as a vital capability, as machine learning algorithms analyze historical data to anticipate equipment failures or maintenance needs, preempting costly downtime and prolonging device lifespan. This paper is considering research in area of machine learning for predictive analysis that are made for blockchain enabled IoT system. Paper has focused on role of ML based predictive analytics and conventional research in related area. Moreover works related to blockchain enabled IoT system and ML based predictive system are focused with their methodology, limitations, outcomes and future scope.

A Critical Review - Use of Ensemble Methods in Intrusion Detection System

Indira P. Joshi, Vijaya K. Shandilya — 2024-03-26

IDSs are essential to the security of contemporary ICT systems. IDSs detect and report attacks, which are frequently examined by administrators tasked with thwarting the assault and reducing damage. As a result, it's critical that the IDS's alerts are as thorough as they can be. In this study paper has offered a multi-layered behavior-based IDS that classifies network using ensemble learning approaches. The ensemble has been built using Decision Trees, NB, SVM and Random Forests, these popular and well-liked models. Our solution is made to rapidly filter away traffic that has been identified as benign without further research in order to speed up system response time, while suspicious events are looked into to produce a more precise categorization. According to experimental setup has discussed on the various public datasets, the system can detect nine forms of high performances across all parameters taken into consideration.

A Comparative Study of Artificial Intelligence and Machine Learning Algorithms for Cybersecurity

Sai Kiran Arcot Ramesh — 2024-03-26

The rapid expansion of cyberspace has been facilitated by a range of innovative networking and computing technologies, including software-defined networking, big data, and fog computing. Currently, cyber security has emerged as a paramount concern in the realm of cyberspace. The security of cyberspace has had significant effects on multiple essential infrastructures. The passive protection approach is no longer effective in safeguarding systems against emerging cyber risks, such as advanced persistent threats and zero-day assaults. So, the main objective of this study is to conduct a thorough examination of different implementations of artificial intelligence in the field of cybersecurity, encompassing activities such as identifying potential risks, responding to security incidents, and utilizing predictive analytics. The methodology employed in this study is qualitative research technique. The study emphasizes the efficacy of AI-powered solutions in strengthening the robustness of contemporary cybersecurity frameworks, based on current case studies and breakthroughs in machine learning algorithms. The paper critically examines the constraints and possible prejudices in AI systems used for cybersecurity, highlighting the significance of responsible AI methodologies. The study will be a contribution to the researchers, practitioners, and policymakers to know about the present condition of artificial intelligence (AI) in cybersecurity. It aims to encourage discussions on the efficient incorporation of AI technologies to tackle the continuously expanding challenges in the field of cyber threats.

Elevating Customer Experiences and Maximizing Profits with Predictable Stockout Prevention Modelling

Sumit Mittal, Priyanka Koushik — 2024-03-26

Preventing stockouts while optimizing revenues is a constant problem for inventory management in retail and supply chain operations. In order to detect stockouts and optimize inventory levels, this study investigates the effectiveness of MLmethods in tackling these problems. Three well-known ML algorithms—Random Forest, GBM, and LSTM—were applied and contrasted using a dataset with 2000 rows and 15 columns that captured various variables linked to inventory management and stockout events. To ascertain how preprocessing methods affected algorithm performance, three different approaches—feature scaling, dimensionality reduction, and no preprocessing—were assessed. The findings show that in terms of accuracy, precision, recall and F1 score, ensemble learning algorithms—in particular, Gradient Boosting and Random Forest—performed better than LSTM. Furthermore, all algorithms performed noticeably better when features were scaled using MinMaxScaler, underscoring the significance of preprocessing in raising model accuracy.

These results add to the body of literature by highlighting the importance of preprocessing approaches in the optimization of inventory management strategies and offering empirical proof of the efficacy of ML algorithms in stockout prevention tasks. Businesses can improve customer happiness, improve inventory management procedures, and reduce financial losses from stockouts by utilizing cutting-edge machine learning techniques. This study highlights how ML-based strategies can spur innovation and enhancement in supply chain and retail operations.

Digital Forensics Investigation Framework Based on the Blockchain, IOT, and Social Networks

Vinod Kumar Uppalapu, Ajay Agarwal — 2024-03-26

Digital forensics involves the identification, preservation, analysis, and presentation of digital evidence to support legal investigations. This paper introduces a novel blockchain-based framework for digital forensics (DF) within the context of Internet of Things (IoT) and social systems. The proposed framework, named IoT forensic chain (IoTFC), capitalizes on the decentralized nature of blockchain technology to address the integrity and provenance challenges of evidence collection across jurisdictional boundaries. By leveraging blockchain's features, IoTFC ensures authenticity, immutability, traceability, resilience, and distributed trust among involved parties. The framework enhances transparency through recorded chains of blocks, covering evidence identification, preservation, analysis, and presentation. This project also presents a secured communication scheme using Blockchain for defense applications, providing privacy through message signing with corresponding private keys.

Cost Effective Analysis of Supplying Essential Food Materials to Indian Armed Forces through Drone Application Versus Conventional Transport by Army Trucks

Gurprit Singh, Ampu Harikrishnan — 2024-03-26

BA stands for cost-benefit analysis. Comparative advantage analysis serves as an economic evaluation technique that quantifies the positive and negative outcomes of a particular intervention or program regarding financial worth. By assigning monetary value to all program outcomes, decision-makers can directly compare the results of various decisions. This research paper provides an overview of the Indian Army drone application, including its motivation, methodology, implementation details, and future scope. The objective of the application is to equip soldiers with the ability to choose the most appropriate drone for payload delivery tasks based on load weight, thus improving operational efficiency and effectiveness. The primary purpose of this application is to shield the tourism and hospitality sectors from the destructive effects of weapons and shelling. The drone also minimizes the impact of sending these pots, arms, and ammunition by heavy vehicles, minimizing the impact of heavy vehicles damaging hilly areas, which are prone to landslides that further damage the soil and surrounding vegetation.

Leading the Way in Efficient Web Content Mining through Advanced Classification and Clustering Techniques

Yogesha T., Thimmaraju S. N. — 2024-03-26

The clustering techniques in online content mining for knowledge discovery is the main topic of the abstract for the article "Clustering Techniques in Knowledge Discovery for Web Content Mining". The application of association rule mining, sequential pattern discovery, and clustering as data mining techniques for knowledge extraction is mentioned.

When the data comes from the online, web mining—the process of obtaining information from web data—is referred to as a subset of knowledge discovery from databases (KDD). A particular kind of web mining called web use mining (WUM) seeks to identify, assess, and make use of hidden knowledge from online data sources. Data from user registration forms, server access logs, user profiles, and transactions are used in web use mining.

It is mentioned that one technique utilized in online content mining for knowledge discovery is clustering algorithms. In the context of online content mining, clustering is the process of assembling comparable data points into groups according to their shared traits or patterns. Clustering may be used to find page sets, page sequences, and page graphs.

The use of text analysis methods for knowledge discovery from unstructured materials, including feature extraction, theme indexing, clustering, and summarization, is also mentioned in the abstract. Press releases, emails, notes, contracts, government reports, and news feeds are just a few of the documents from which valuable information may be extracted thanks to these strategies.

An overview of the use of clustering algorithms in knowledge discovery for online content mining is given in the abstract overall. It highlights the use of text analysis tools to extract knowledge from unstructured documents and the clustering approach in online use mining.

Enhancing Concept Drift Classification in Computer Networks with Artificial Intelligence through NCDC-DM: A Novel Approach Utilizing Diversity Measure

2024-03-26

nowadays data stream mining has been essential area for research work which has been getting waste focus because it was utilized along a huge counts of applications, like telecommunication, networks of sensors & banking. Important issue was effecting mining of data stream was concept drift. When contact between target variable & input data modifications at this time. In last decade there are many classification techniques of concept drift was proposed, that either getting problem of high cost along conditions regards memory either run time either, that was not quack along conditions of classification speed. This paper proposes a technique which known as Novel Concept Drift Classification utilizing Diversity Measure (NCDC-DM), along reduction of less memory & less time it is reacting quickly. Under proposed system there is collaboration between disagreement measure & diversity measure, which known through static learning along scenarios of streaming utilizing test of page & utilizes these calculations comparatively along classification technique of ten drift utilizing various scenarios of drift. Outputs of research shows that proposed technique most efficient & it has capability of faster classification concept drift & it’s compared with existing ADASYN, EACD, HLFR methods.

Advancing Network Lifetime in Wireless Sensor Networks through Localization Techniques: A Perspective from Computer Networks

2024-03-26

the primary task of a Wireless Sensor Networks (WSNs) is to sense the environment around, and sends the information back. The sensor nodes need to be of less size, low on power consumption which substantially constrained the computational capacity of these nodes. So any computational task involving these nodes must be very power efficient so that the duration of the deployment can be increased. Localization Techniques in WSNs have been created to identify unknown sensor node position information. This is a fundamental need in many different applications, hence it was necessary to develop these techniques. This is a general rule. In this work, our primary emphasis is on exploiting the information provided by anchor nodes to more accurately estimate the positions of sensors whose locations are unknown while simultaneously reducing the amount of power that is required to do so. It is as yet a troublesome issue to locate a precise and efficient node location computation algorithm in sensor networks. In this paper, we proposed a distributed technique for localization of sensor nodes utilizing couple of mobile reference nodes. When compared to two current energy-efficient clustering and localization methods—the ECGAL and the CLOCK-Localization Approach—the study outputs reveal that the suggested methodology is the most efficient and has the capacity of being speedier.

Neurocomputing assisted Consensus Based Web-of-Service Software Design Optimization: A Fault-Resilient Reusability Prediction Approach

Prakash V. Parande, M. K. Banga — 2024-03-26

In this paper, a state-of-art new neuro-computing assisted consensus-based ensemble model was developed for Web-of-Service (WoS) software reusability prediction. In order to achieve higher accuracy with reliability of prediction, the proposed model made enhancement in both data-model as well as classifier-model. More specifically, it applied WSDL-CKJM tool to extract object-oriented-programming (OOP) metrics, which were subsequently processed using univariate logistic regression-based feature extraction followed by sub-sampling method. In the proposed reusability prediction model, to alleviate data or class-imbalance and skewness problem, three different sub-sampling methods were applied including up-sampling, down-sampling and SMOTE sampling. Once obtaining the differently sampled data with the confidence interval of 95%, it was amalgamated together to give rise a composite feature vector pertaining to WMC, CBO, DIT, LCOM, NOC, and RFC OOP-CK metrics, characterizing structural features of the software program. Subsequently, to alleviate computational overhead Wilcoxon Rank Sum Test (WRST) was applied, which retained the most suitable feature set towards reusability prediction. To alleviate the problem of convergence and over-fitting, Min-Max normalization was performed over the selected feature set. Thus, the normalized input features were processed for two-class classification using the proposed neuro-computing assisted homogenous ensemble model. Noticeably, being homogenous ensemble structure, we used ANN variants with gradient descent (GD), radial basis function (RBF), Levenberg Marquardt (LM) and probabilistic neural network (PNN) as base classifiers. The aforesaid base-classifiers helped in estimating the consensus to make each-class classification, where the proposed consensus-based classification model achieved superior accuracy (96.57%), precision (0.94) and recall (0.99), signifying its robustness over the classical standalone classifiers.

Women Vulnerability Index (WVI): Multi Criteria Decision Making Approach

Seema Aggarwal, Geeta Aggarwal, Manisha Bansal — 2024-03-26

Crime against women, a never-ending issue is a sad reality that demands focused attention. The occurrence of crimes in different states in India varies a lot. Multi-Criteria Decision Making (MCDM) method, called TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) is applied on real occurrences of crime to develop women vulnerability index (WVI). This index measures the susceptibility of women to crime in any region of India. This marks the first instance of applying MCDM technique (TOPSIS) to derive such an index for crime against women. The index will equip the law enforcing agencies and various NGOs to assess the susceptibility of Indian women in different regions and take appropriate action for mitigation of such crimes to create a safe environment for women. We find that states like Mizoram, Nagaland, Sikkim in northeast India and Lakshadweep Islands in southern India have very low values of the index and are the safest places for women. On the other hand, Uttar Pradesh, Delhi, Haryana, Rajasthan, and Bihar are Indian states where women are most susceptible to crime having very high values of WVI.

A Novel Algorithm for Breast Cancer Detection: An Overview

M. Ida Rose, Mohan Kumar — 2024-03-26

Breast cancer stands as a significant global health issue impacting millions of women. Detecting it early is pivotal for enhancing the prognosis and survival rates of affected individuals. Recent years have witnessed a surge in research dedicated to crafting innovative algorithms and employing machine learning techniques to facilitate early breast cancer diagnosis. These cutting-edge approaches harness diverse imaging modalities and computational methods to elevate accuracy and efficiency. This study introduces a distinctive algorithm designed to forecast the percentage of breast cancer through the analysis of mammogram images. The algorithm incorporates a variety of techniques to enhance its accuracy and overall performance. These methodologies include data augmentation, dropout layers, the RMSprop optimizer with learning rate decay, sparse categorical crossen tropy loss, and an increased number of training epochs. Data augmentation is employed to generate a diverse set of training examples by applying random transformations to the images. This process enriches the model's ability to generalize to unseen data. Dropout layers, strategically placed after Conv2D and Dense layers, serve as a preventive measure against overfitting, thereby improving the model's generalization capabilities. The use of the RMSprop optimizer with learning rate decay offers precise control over the learning rate during training, enabling faster convergence and potentially reaching a more optimal solution. A thorough analysis of these features allows the algorithm to predict the probability and percentage of breast cancer in a given patient. The results demonstrate a strong correlation between the algorithm's predictions and the actual percentage of breast cancer, highlighting its accuracy and reliability. In an extensive cohort study, the algorithm exhibited exceptional accuracy in predicting the percentage of breast cancer, surpassing traditional methods in both sensitivity and specificity. The introduction of this algorithm holds great promise in supporting healthcare professionals with the early detection and diagnosis of breast cancer. It anticipates advancements in patient outcomes and the formulation of personalized treatment strategies.Its multifaceted approach, integrating various techniques, positions it as a robust solution in the ongoing efforts to enhance breast cancer diagnosis and patient care.

Integration of Ethereum Blockchain with Cloud Computing for Secure Healthcare Data Management System

Ginavanee A., S. Prasanna — 2024-03-26

Effective management of health data is critical in this age of digital change. In order to solve the issues with safe healthcare data systems, this paper suggests a brand-new Hybrid Healthcare Data Management System (HDMS) that seamlessly combines blockchain and cloud computing technology. The solution guarantees the safe storage and effective administration of health data by utilizing the scalability and flexibility of cloud computing. The Ethereum blockchain, which offers immutability through smart contracts for data integrity assurance, is used to improve security. Important components of the suggested approach include Blockchain Anchoring, which uses distinct hash references on Ethereum to ensure data integrity, Google Cloud Integration for scalable storage and immediate data access, and compliance with Health Level 7 (HL7) formatting criteria for medical data storage. Robust privacy safeguards include cutting-edge methods like decentralized identifiers (DIDs) and homomorphic encryption, which guarantee secure computations on encrypted data and trustworthy identification for allowed access. IPFS file storage is used by the system to improve security by providing redundancy and resilience. By utilizing blockchain's immutability and cloud storage's distributed architecture, HDMS demonstrates resilience against disturbances. Results of the proposed method is implemented in Python Software. The Homomorphic approach that has been suggested continuously beats Optimized Blowfish Algorithm (OBA) in terms of encryption and decryption times. The improvement in encryption is between 3000ms (10 kb) and 1350ms (40 kb). The enhancement demonstrates improved efficiency across data sizes, ranging from 4000ms (10 kb) to 3350ms (40 kb) in decryption. The goal of HDMS is to create a future in which patient outcomes are enhanced by the appropriate handling, storage, analysis, and use of medical data.

Attendance Tracking with Perception Detection using Recurrent Neural Network

Glodelyn D. Ocfemia, Karren V. De Lara — 2024-03-26

The online class environment has become prevalent in education since the Covid-19 pandemic. It offers students and educators the flexibility to access online education from various locations. Despite its challenges, the online class environment has become essential for providing accessible education, especially during the times of disruptions. There are still educators that conduct attendance checking manually so the student’s attendance and attentiveness during online classes is a big challenge for teachers. The researcher developed a system that can detect body motion and face expression in an online class setup. Attendance tracking with perception detection with the use of Recurrent Neural Network is suggested to detect body motion and face expression, capture and then, store the students’ attendance in the system. The system will use real-time detecting the body motion and face expression whether the students is attentive or not are being captured to store the student attendance during online classes. The body motion will detect the body and the facial expression to capture and track when the camera is on. The system incorporates a notification within the system for faculty regarding inattentive students. Also, the system developed as a separate web application that is compatible or complementary with the existing setup of online classes.

Analysis Of Skin Cancer Detection Using Svm & Resnet-50

Rafik Ahmad, Kalyan Achariya — 2024-03-26

The paper utilizes machine learning algorithms that incorporate Support Vector Machines (SVM) and Resnet-50, in detecting skin cancer from dermoscopy images. The study evaluates the performance of both models using accuracy, confusion matrix, graphs, and Receiver Operating Characteristics (ROC) to determine which model is more effective in skin cancer detection. Previous studies suggest that Resnet-50 outperforms SVM in terms of detection accuracy. Therefore, this paper also demonstrates the potential of combining both models to improve skin cancer detection accuracy. The outcomes of this study hold substantial inference for the field of clinical practice. By using computer-aided diagnosis (CAD) systems, clinicians can make more accurate diagnoses of skin cancer, reducing interobserver variability and improving objectivity. This research underscores the capacity of machine learning models to transform the aspect of skin cancer diagnosis and treatment, ultimately leading to enhanced patient outcomes. The abstract offers valuable perspectives on the efficiency of machine learning models in the realm of skin cancer detection, rendering it a valuable point of reference for researchers and clinicians exploring the usage of machine learning canon in this domain.

Recognition and Classification of Skin Cancer using Deep Learning

Rafik Ahmad, Kalyan Achariya — 2024-03-26

Melanoma, a type of skin malignant growth, is a developing problem in the clinical world. This malignant growth, starting in the epidermal layer in cells which gives color to the skin called melanocytes, has metastatic inclinations with high prospects of arriving at nerves and bones and causing lethally unfavorable impacts. Melanoma's apparent side effects are injuries on cutaneous surfaces with trademark properties which are key determinants for specialists to separate between a harmless or dangerous sore. Subsequently, an extremely huge advance to lessen the death pace of Melanoma is early analysis with high precision during the essential improvement time of sore Clinical pictures of such skin abnormalities are analyzed utilizing the painless act of dermoscopy. Dermoscopic pictures are gotten through Medical Imaging Procedures anyway their appraisal was physical and relied vigorously upon the dermatologist's comprehension. Presently the central technique utilized for assessment of a sore is ABCD measures which set norms for four boundaries of an injury via Asymmetry, Border Irregularity, Colour Pigmentation and Diameter (>6mm). Injuries satisfying ABCD measures need quick master consideration. Endeavors for reproducing ABCD models on mechanized frameworks utilizing techniques for picture handling for symptomatic precision and speed have been made before. Any way central issues with these modalities incorporate uncertainty inside human comprehension, goal restrictions, bending and unfortunate differentiation, algorithmic mistake of doling out the same mathematical qualities to divergent sore boundaries and impediments of ghastly strategies by the powerlessness of acquiring exact recurrence content of the injury's boundary. Our undertaking utilizes Keras and Matplotlib library of Python to prepare a model on disease order.

Bird monitoring intelligence: Integrating Thermal UAV Imagery and Deep Learning Tools

2024-03-26

Birds are excellent indicators of biodiversity and due to their selective association, are ideal for providing insights into the diversity of vegetation, insects, and aquatic life. Bird census, therefore, is an important tool for ecological monitoring. Birds, however, particularly migratory birds, often flock together in large numbers and bird count estimation of such congregations are herculean tasks, subject to large error margins. Computer vision tasks, such as object detection, tracking, and counting, immensely aid environmental monitoring. Despite many innovative techniques, there is still a lot scope for simplifying and improvising the count estimation process, especially through employing technology and artificial intelligences (AI). The integration of AI into drones for on-the-fly problem-solving is an evolving trend. This paper endeavors to offer a comprehensive compilation of potential studies on wildlife using low-altitude UAVs equipped with thermal sensor datasets found in the literature. Further, we tested the ability of a thermal drone to identify and count water birds in a fresh water habitat. The Unmanned Aerial System with an optical and thermal sensor was integrated with widely accepted detection models such as Detection Transformer, Yolo V7 and Yolo V8 to delineate and count the birds. Thermal imagery was found to be excellent in highlighting birds as bright/hot pixels especially against the cooler waterbody. Among the models, DETR achieved the highest precision score of 91.4%, followed closely by YOLOv8 with a precision score of 84.1%. Additionally, DETR exhibited a notable mAP of 89.2%, demonstrating its efficacy in object detection tasks. Interestingly thermal images are also effective in detecting birds even through canopy that otherwise camouflage well in vegetation. The birds didn't show much response to the presence of UAS particularly at late hours of the day. There is a huge scope of applications and research in the field of ecology. Our study illustrates how UAS, thermal imagery, and automated detection algorithms can be combined to efficiently detect and count birds, thereby offering a critical solution towards population count estimation essential for wildlife management.

Next-Generation Spatial Data Management Leveraging Spatial Databases and Blockchain in Cloud Data Architectures

2024-03-26

GeoChainDB, a breakthrough geographical data management platform, solves data security, scalability, and openness issues with spatial databases, blockchain, and cloud architectures. The latest technology allows successful geographical data management in many contexts. SpatialDataIngestion inputs data fast and precisely, BlockchainConsensus secures agreements, and CloudScalability enables GeoChainDB cloud-based administration flexibility. Flowcharts and equations explain each procedure. SpatialDataIngestion effectively imports spatial data using rates and a validation score. BlockchainConsensus finds consensus, calculates consensus time, and checks security score for transaction integrity. CloudScalability quantifies and assesses resource utilization to scale geographic data management in cloud systems. These algorithms have flowcharts that demonstrate their ability to handle geographical data, secure blockchain consensus, and cloud scalability. Two more tables compare speed and economy to previous approaches. The results indicates that GeoChainDB outperforms earlier techniques across several criteria. Better data security, scale, and openness make GeoChainDB a solid spatial data management choice. Its mathematics and graphics make it a better spatial data management platform than prior techniques.

A Novel Zipper Logic Based Hybrid 1-Bit Full Adder Circuit Design

Guduru Devi Charan, Thumbur Gowri — 2024-03-26

Zipper logic is a design that integrates many logic types or approaches to create an efficient and optimal circuit. A 1-bit Full Adder (FA) is a basic building block in digital circuits that adds two 1-bit binary inputs (A and B) plus a Carry Input (Cin) to create a sum output (S) and a Carry Output (Cout). This study examined the concept of a zipper logic-based hybrid 1-bit FA circuit design, which pertains to a particular method of designing a digital circuit that performs the addition operation on two 1-bit binary integers (bits). The performance parameters of the hybrid circuit are assessed and compared to standard 1-bit FA designs, revealing considerable improvements in latency and energy usage. The suggested FA cell's performance has been analyzed using a Cadence simulator at the 16 nm production node, where it was compared to that of existing FAs throughout a supply voltage spectrum of 0.4 to 1.0 V. When compared to the other adder, the suggested adder improved Average Power Consumption (ADP) by 48.9% and Power Delay Product (PDP) by 66.7% when conducting at 0.8V.

Diabetes Care: A Machine Learning Based Review Under Supervision and without Supervision

2024-03-26

Diabetes is a persistent metabolic condition that affects millions of people globally. The effective management of diabetes care is crucial in order to prevent complications and improve patient outcomes. Recent years have seen a substantial increase in the use of machine learning techniques in the field of healthcare, especially the treatment of diabetes. This review seeks to offer a thorough examination of machine learning techniques used in diabetes treatment, both supervised and unsupervised. Algorithms for supervised machine learning have been widely used for a variety of diabetes care activities, including risk assessment, diagnosis, and medication recommendation. These algorithms utilize labelled data to train predictive models, allowing for accurate identification of high-risk individuals, early detection of diabetes, and personalized treatment plans. In particular, support vector machines, random forests, and synthetic neural networks have produced promising outcomes in these fields of contrast, unsupervised machine learning techniques have been used for pattern identification and exploratory analysis of big datasets without specified labels. The identification of patient subgroups based on shared traits using clustering techniques like k-means and hierarchical clustering has enabled personalised therapies and precision medicine approaches in the treatment of diabetes. Principal component analysis and t-distributed stochastic neighbour embedding are two examples of dimensionality reduction techniques that have been useful in visualising complex data and revealing hidden relationships. This review also discusses the challenges and limitations associated with the application of machine learning in diabetes care. Issues such as data quality, interpretability, and generalizability of models are addressed, highlighting the importance of addressing these concerns for successful implementation in clinical practice.

In conclusion, the integration of supervised and unsupervised machine learning techniques holds great potential in improving diabetes care. These methods provide valuable insights into risk assessment, diagnosis, treatment, and patient stratification. Nonetheless, further research and collaboration between data scientists, clinicians, and researchers are necessary to address the challenges and enhance the translation of machine learning algorithms into real-world clinical settings.

Internet of Things Security and Privacy: A Systematic Investigation

S. Shiva Prakash, M. P.Vani — 2024-03-26

The Internet of Things (IoT) paradigm shift is one of the most remarkable phenomena of recent times.Many security issues are triggered when disparate IoT devices are combined with the traditional Internet. This is because conventional internet connection methods were never intended to accommodate IoT. As a result, there are now countless ways in which IoT-enabled equipment could be compromised. This assessment of the literature centred on the primary concerns regarding the safety of IoT. The adaption of machine learning techniques in IoT, as well as its layered architecture, protocols for communication, and energy-efficient data routing strategies, were all subjects of our research. In addition to discussing the current attacks, hazards, and state-of-the-art remedies, we lay out a road map for meeting the security needs of the Internet of Things. We also compile a table of IoT security issues and a map of published remedies. We conclude that an attacker can compromise an IoT device, infrastructure, or network based on the results of this study. As a result of the information gathered in this poll, researchers are more motivated than ever to create a foolproof intrusion detection system for the Internet of Things (IoT).

Designing a Framework for Developing an Adaptive Information Retrieval System that Personalizes Information

B. Sangamithra, Manjunathswamy B. E., Sunil Kumar M. — 2024-03-26

As a result of advancements in internet technology, more and more people are turning to the World Wide Web as their primary source of information and education. The academic and business communities have shown considerable interest in personalized search due to its potential to improve the effectiveness of Web searches. In comparison to a standard web search, customized search returns results that are tailored to the individual. Each user of a personalized Internet search will see a unique set of search results tailored to their own set of interests, tastes, and information needs in response to any given query. Unfortunately, the current personalized search methods fall short of fully meeting the needs of the particular user, as they do not take into account either the user's most recent preferences nor the interests of other users. With the rise of Personalized Search, however, comes a new challenge: users' reluctance to reveal sensitive information about themselves during searches. The most common search engines are made with everyone in mind, rather than a specific user in mind; as a result, the results they return for a given query are generic, rather than tailored to the individual user. Numerous algorithms exist to swiftly analyze user preferences and return relevant search results via a personalized web search;. Examples of applications for such algorithms include user tracking, link analysis, textual analysis, and collaborative online search. This paper mainly designs a framework by an adaptive information retrieval system which presents more appropriate information for users. The experimental results show that our proposed framework reduces the search time and improves the efficiency of web search.

Integrated Approach for Crop Yield Prediction in Telangana Region Using Ensemble Techniques and ARIMA Model

P. Sowmya, A. V. Krishna Prasad — 2024-03-26

This paper presents an integrated methodology for accurate and comprehensive crop yield prediction in the Telangana region, spanning the years 1966 to 2030. Leveraging an ensemble approach, our model combines the strengths of Random Forest Regressor at both the state and district levels, providing granular predictions for each administrative unit. Additionally, we employ an ARIMA model to forecast key meteorological and soil parameters from 2021 to 2030. The ensemble predictions are then integrated with historical data, resulting in a holistic forecast for crop yield. The methodology addresses data sparsity by replacing zeros with mean values, enhancing the reliability of predictions. The proposed approach is validated using robust metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared, demonstrating the robustness and accuracy of the model. The study contributes to the field of precision agriculture, offering insights into the complex dynamics influencing crop yield and providing a valuable tool for sustainable planning in the Telangana region.

Deep Learning Based Facial Emotion Recognition for Analysing the Effectiveness of Online Class

Sophiya Mathews, D. John Aravindhar — 2024-03-26

Online classes break down barriers of distance and time, allowing students from different geographical locations and backgrounds to access quality education. However, monitoring student engagement and emotional well-being during online classes presents a unique challenge. This study aims to analyze the facial expressions of students during online classes, in order to assess their emotional states and evaluate the performance of a fine-tuned MobileNet V2 architecture. To conduct this study, we utilized the CK+ dataset, which consists of labeled facial expressions captured in controlled laboratory settings. To specifically identify the emotions shown by students during online classes, the MobileNet V2 model is first pre-trained on ImageNet, a large-scale picture classification dataset, and then refined on the CK+ dataset. Preprocessing techniques such as image augmentation and normalization are applied to enhance the model's generalization capability. Before fine-tuning, the pre-trained model achieved moderate level of performance. After fine-tuning, the performance of the model achieved higher accuracy of 98.40% compared to the base model, indicating its enhanced ability to detect and classify facial emotions during online classes. By leveraging deep learning-based tools like the proposed model, educators can gain valuable real-time feedback on the effectiveness of their online teaching methods and make data-driven decisions to optimize the learning experience for their students.

Automated Detection of Pulmonary Pathologies through Deep Learning in Lung Ultrasound

Anjelin Genifer Edward Thomas, J. Shiny Duela — 2024-03-26

The current surge in newly reported pulmonary diseases and the possibility of further epidemics necessitate the immediate development of a novel Deep Learning (DL) model to facilitate early diagnosis and treatment. Lung ultrasound (LUS) has the potential to detect symptoms of a pulmonary infection, based on growing evidence from various nations. Several characteristics of ultrasonic imaging make it well-suited for routine use: Small hand-held systems, unlike X-ray or computed tomography (CT) equipment, are easier to clean because they are encased in a protective sheath. LUS, on the other hand, enables patient triage in settings other than hospitals, such as tents or homes, and it can detect lung activity during the early stages of the disease while also monitoring affected patients at the bedside on a daily basis. This review paper discusses the potential applications of LUS imaging for disease segmentation and categorization. The paper investigates the open-access LUS dataset and examines image processing algorithms that could increase pulmonary disease detection and segmentation accuracy. We also discuss the many segmentation strategies available for LUS images. Next, we present the currently available DL approaches for LUS image categorization. This survey can be extremely beneficial to researchers struggling with disease diagnosis using LUS images, providing excellent advice on how to proceed with their investigation and determine the source of the problem.

Enhancing Efficiency in Cloud Computing Entails Optimizing Resource Apportionment Through the Utilization of the Shuffled Frog-Leaping Algorithm (SFLA) and Firefly Algorithm

Namrata H. Patadiya, Nirav V. Bhatt — 2024-03-26

The imperative role of 'cloud computing' in modern technology brings attention to Resource apportionment as a pivotal facet. This paper introduces a Hybridized Optimization algorithm that combines the effectiveness of the 'Shuffled Frog Leaping Algorithm' (SFLA) and the 'Firefly Algorithm.' This innovative approach overcomes limitations seen in current works like the HABCCS algorithm, GTS algorithm task, and the krill herd algorithm, while amalgamating the unique features of both SFLA and the Firefly Algorithm. Within this methodology, the SFLA section oversees initial steps, encompassing the initialization of request size, request generation, estimation of SFLA's fitness value, sorting, division, and evaluation of user requests. SFLA is recognized for its rapid convergence and straightforward implementation, boasting the capability for global optimization and widespread utilization across diverse domains. Concurrently, the Firefly Algorithm takes on pivotal operations such as initialization, request generation, fitness function evaluation, modification, and the assessment of new solutions. The Firefly Algorithm is characterized by its ease of evaluation and suitability for complex situations, providing a notable advantage. In this system, the evaluation of request speed and sizes plays a critical role in Resource apportionment on the server side, contributing to reduced computation times. Experimental results substantiate the efficacy of this hybrid approach, illustrating its superior performance in comparison to additional similar technique.

Autoencoder-Boosted Lightweight Dense Net for Dimensionality Reduction and DOS Attack Classification in WSN

Sarkunavathi A., Srinivasan V., Ramalingam M. — 2024-03-26

Wireless Sensor Networks (WSNs) are liable to Denial of Service (DoS) attacks, which can be easily executed in this context. This study presents a comparative analysis of five prominent deep learning architectures, namely AlexNet, VGGNet, ResNet, DenseNet, and Lightweight DenseNet, for their efficacy in classifying Denial of Service (DoS) attacks in Wireless Sensor Networks (WSNs). The evaluation is conducted using labeled instances of different types of DoS attacks from the WSN-DS and IOTID20 datasets. Various evaluation metrics including F1-score, recall, precision and accuracy computational efficiency are employed to discern the suitability of these architectures for real-time WSN applications. Experimental results from training and testing on the WSN-DS and IOTID20 datasets provide insights into the performance of each architecture, aiding in the selection of optimal models for DoS attack classification in WSNs.

A Novel Weight Based Interest Forwarding Protocol for Information Centric Networking

Krishna Delvadia — 2024-03-26

Information centric network (ICN) is a new communication paradigm that is introduced to satisfy the needs of internet users in context of throughput and delay. Content request routing is an important research domain of content centric network. If request is routed efficiently within network, then retrieval of desired content is possible in least duration with less overhead. This paper introduces a weight-based interest forwarding strategy that aims to route interest message towards content router (CR) having maximum likelihood of having desired data. This can significantly contribute in reduction of latency and overhead. The protocol exploits three different parameters namely interest packet forwarding ratio, size of node’s pending interest table (PIT) and count of data messages produced by router to take decision for interest packet forwarding. The experimental analysis of proposed strategy is done inside ndnSIM 2.0. The performance testing of state-of-the-art caching mechanisms with and without inclusion of proposed protocol in context of data discovery delay, overhead and content store (CS) hit ratio. We have also compared the integrated variants of protocols against recent existing forwarding protocols. The extensive simulation study proves that the coupling of proposed mechanism to existing mechanisms remarkably enhances the performance by 10-42%.

Use of Machine Learning for Personalized Care for Persons with Disabilities: Ethical and Privacy Issues

Guillermo V. Red Jr., Thelma D. Palaoag — 2024-03-26

Machine learning remains the future of care personalization, especially in caring for persons with disabilities. However, critical ethical and privacy challenges arise, which potentially derail the deployment of machine learning technologies. This research has explored these challenges, considered possibility of machine learning being used to resolve the challenges, and examined the policy and governance frameworks that could be used to address the problem. The study adopted a scoping review method, where a scoping review process was conducted for each of the three research questions. This approach made it possible to offer an in-depth assessment of all aspects of the research problem. The findings support the available literature by establishing that critical privacy and ethical challenges exist. These challenges fall under three main categories: the technology, practice, and data. Like any other technology, machine learning is vulnerable to malicious activity and its use could breach patient privacy. Since it involves handling patient data, there is a possibility that the handling of the data itself poses privacy and ethical risks.

Improving Fog Gateway with Novel Metaheuristic-Driven AI Technique for Lessening the Delay and Energy Measures

Pooja Grover, Swati Singh, Beemkumar Nagappan, Amandeep Gill, Abhiraj Malhotra — 2024-03-26

The increasing need for quick data transmission and energy efficiency at the edge of the network has led to the development of a technology known as fog computing. Fog gateways are crucial elements in the architecture, as they offer computational capacity and enhance accessibility to end users and Internet of Things (IoT) devices for data-absorbing tasks. Improving latency and improving energy efficiency at the Fog Gateway remains a significant challenge. This research proposes a framework utilizing metaheuristic-driven artificial intelligence (AI) techniques to address the problem. This work introduces an innovative snow ablation search-driven catboost (SAS-CB) approach for identifying computational demands. The information from the IoT-driven fog computing system is utilized to develop the proposed SAS-CB method. We utilized sensors to gather environmental information for this research. Further feature selection is carried out utilizing the snow ablation optimization (SAO) technique to decrease the misinterpretation rate of the CB technique. The proposed method is implemented on a Python platform and evaluated based on several metrics such as utilization of energy (9W), latency (20s), and accuracy (90.45%). The experimental evidence indicates that the suggested solution outperformed existing methods in enhancing the fog gateway with favorable energy and latency parameters.

Novel prediction mechanism for Attack Prevention in Fiber-Optical Networks using AI-based SDN

2024-03-26

Fiber-optical networks enhance communication by delivering data through light signals, which leads to fast and secure communication. Technological advancements provide difficulties, such as the exposure of Artifiical Intelligence (AI) based Software-Defined Networking (SDN) to attacks of distributed denial-of-service (DDoS). The integration of fiber-optical networks and AI-powered SDN highlights the essential requirement for comprehensive cyber security regulations to protect the integrity of current communication infrastructure. In this research, we developed an innovative strategy named Sea Lion fine-tuned Long Short-Term Memory (SL-FLSTM) to predict the attacks of DDoS in fiber-optical networks. Initially, we gathered a dataset which includes fiber optic network communication traffic with various types of DDoS attacks, to train our proposed approach. Our suggested SL-FLSTM incorporates insights from Sea Lion (SL) behavior to improve sequential data processing; it integrates bio-inspired modifications into the LSTM architecture, improving long-term dependency modeling. Min-max normalization algorithm is used to pre-process the gathered raw data, for enhancing the quality of the data. The suggested approach is implemented in Python software. The result evaluation phase is performed with multiple parameters including recall (98.1%), precision (98.2%), F1 score (98.3%) and accuracy (98.4%) to evaluate the suggested SL-FLSTM approach with other conventional methodologies. The experimental results demonstrate that the proposed SL-FLSTM approach performed better than other existing approaches in predicting DDoS attacks in fiber-optical networks.

Developing An Innovative Image Processing Model For Computer Networks through Optimized K-Nearest Neighbour Algorithm

2024-03-26

Image processing is the process of enhancing or extracting information from images. It includes a wide range of methods, including segmentation and pattern recognition. In the context of computer networks, image processing is critical for enhancing data transfer and communication efficiency. The integration of image processing with computer networks improves the overall efficiency of visual information collaboration, which leads to innovations in various domains. In this research, we developed a novel machine learning-based image data processing model for computer networks named Red Deer optimized Adaptive K-Nearest Neighbour (RD-AKNN). We gathered a dataset that includes various types of image data to train our proposed approach for image processing. The data cleaning process is performed to reduce the redundancy, Global Contrast Normalization (GCN) algorithm is utilized to pre-process the gathered raw data. Red Deer Optimization (RDO) is employed to enhance the crucial characteristics of the suggested AKNN architecture for developing an innovative mage data processing model in computer networks. We implemented our proposed methodology in Python software. The finding analysis phase is performed with various metrics such as recall (97.5%), accuracy (97.2%), F1-score (98.3%) and precision (98.1%) to evaluate the proposed algorithm with other conventional methodologies. The experimental results demonstrate that the proposed RD-AKNN approach performed better than other conventional approaches for enhanced image data processing in computer networks.

Enhancing Cloud Security: Artificial Intelligence-based Data Classification Model for Cloud Computing

2024-03-26

Cloud computing (CC) is the Internet-based delivery of computer services, including data retention, processing and programmers. This permits users to develop and improve their electronic devices by providing them with instant utilization of communal assets. Categorizing data using CC is important due to organizes and protects information according to its level of sensitivity. By developing Intelligent Rat Swarm Optimized Adaptive Boosting (IRO-Adaboost), an innovative AI-based data classification approach, we hope to enhance CC environments' security. In order to train our proposed data categorization method, that we initially gathered a collection of data involving various types of data from numerous organizations. The Box-Cox Transformation (BCT) procedures are used for processing the raw data that has been obtained. We employed the Term Frequency-Inverse Document Frequency (TF-IDF) method for extracting useful features for the data that is analyzed. Our suggested approach uses swarm intelligence based on rat behaviour to enhance the performance of the Adaboost algorithm. To evaluate the proposed IRO-Adaboost technique to different standard methods, a number of metrics are employed in the outcome assessment stage, including sensitivity (92%), accuracy (96%), False Negative Rate (FNR-0.2064), False Positive Rate (FPR-0.05), and specificity (95%). The experiment's findings indicate that the recommended IRO-Adaboost strategy worked better than other conventional approaches to increase security in a cloud computing environment.

Optimized Traffic Classification System for Software-Defined Networking using a Deep Learning-based Approach

Trapty Agarwal, Soumya K., Manish Nagpal, Karishma Desai, Krishna Nandan — 2024-03-26

Software-Defined Networking (SDN) increases scalability and flexibility of network administration by eliminating the control as well as data planes. SDN improves network management by doing away with the control and data planes, consequential in increased scalability and flexibility. A traffic classification system for SDN improves network efficiency by classifying the data flows. Quality of Services (QoS) enhances and optimizes use of resources with flexible adaptation to changing network requirements. To create an optimal traffic classification system for SDN, we proposed a novel Deep Learning (DL) approach called Lightning Search fine-tuned Generative Adversarial Networks (LS-GAN). We collected a dataset comprising several kinds of network traffic logs to train the suggested methodology. The obtained raw data is pre-processed using the Unit Vector Transformation (UVT) technique. Kernel Principal Component Analysis (K-PCA) is used with the processed data to determine the key features. The LS-GAN approach combines the potent capabilities of Generative Adversarial Networks (GANs) with blazingly quick search algorithms. The system can effectively and precisely detect different kinds of network traffic inside SDN designs by combining these methods. The proposed LS-GAN obtained a Precision (96.2%), Accuracy (98.3%), Recall (97.3%) and F1-score (98.6%). The experimental outcome show that the suggested LS-GAN approach performed better than existing approaches in SDN infrastructure for increased traffic classification.

AI-Powered Encryption: Innovative Approach for Malware Intrusions in File-Sharing Networks

2024-03-26

Malware intrusions in file-sharing networks are a prevalent problem, compromising the integrity and security of encrypted data. These disguised attacks took the utilization of flaws in network designs, spreading intentionally and corrupting data shared by users. The dynamic and distributed organization of file-sharing networks complicates the detection and prevention of such attacks. In this study, we developed an optimized encryption model named Adaptive Emperor Penguin tuned Bayesian Belief Networks (AEP-BBN) for enhancing the prediction of malware activities in file-sharing networks. Initially, we gathered a dataset that includes infected shared files from the organizations to train our proposed prediction model. A robust Scaling (RS) algorithm is employed to pre-process the gathered raw data, to improve the quality of the data. We extracted significant features from the pre-processed data using Recursive Feature Elimination (RFE). Adaptive Emperor Penguin Optimization (AEPO) is used to enhance the primary features of the suggested BBN architecture. The recommended approach has been implemented in Python software. The result assessment phase is performed with numerous metrics such as recall, precision, f1 score and accuracy to evaluate the suggested AEP-BBN approach with other conventional approaches. The outcomes of the experiments demonstrate that the proposed AEP-BBN approach performed better than other existing approaches for enhancing the prediction of malware activities in file-sharing networks.

Securing the Cloud Storage Using Novel Machine Learning-based Intrusion Detection System

2024-03-26

Intrusion detection is an approach for detecting unauthorized access or malicious behavior inside a system. When applied to cloud storage, it entails monitoring and analyzingnetwork activity, system records and consumer activity to identify any possible security vulnerabilities or abnormalities. This proactive method assists in securing sensitive data and the integrity of cloud-based infrastructure, ensuring persistent security against cyber-attacks. In this study, we produced an intelligent intrusion detection algorithm named Gorilla Troops fine-tuned Modified Random Forest (GTO-MRF) for enhancing security in cloud storage. Initially, we collected a dataset comprising simulated cloud-based intrusion scenarios and a variety of attack types. The proposed model was evaluated using this diverse dataset to enhance security measures. UnitVector Transformation (UVT) algorithm is employed to pre-process the gathered raw data. We extracted primary features from the pre-processed data using Kernel Principal Component Analysis (KPCA). The Gorilla Troops Optimization (GTO) approach improves the approach by adjusting the tree architectures and feature importance weights. We implemented the proposed model in software. The result evaluation phase is performed with multiple metrics such as training time, False alarm rate and encryption time to evaluate the suggested GT-MRF approach. We conducted a comparison analysis with other conventional methodologies. The experimental results illustrate that the proposed GT-MRF approach performed better than other conventional approaches for enhanced intrusion detection models in cloud storage.

PTFIC - Patient Health Tracking through Fog Enabled Internet of Things Network Using Optimized Classifier

Bibi Ameena, Loganathan R. — 2024-03-26

The network creates a lot of data, which may be saved and processed by a cloud system, though Internet of Things (IoT) technology reinforced with a fog-enabled cloud computing system to monitor remotely. The data transit between the user and the cloud is processed to generate the communication, internet service, and cloud storage function. To facilitate remote patient monitoring, we suggested PTFIC - patient health tracking over a fog-enabled Internet of Things network utilising an optimised confidence classifier and a fog layer at the gateway. The fog layer oversees patient data validation and transmission by utilising the classifier to analyse significant events by determining the health state.

Improving Medical Imaging Diagnostics by Utilizing Compression Techniques and Implementing FPGA Acceleration with Vivado HLS

2024-03-26

this paper aims to provide efficient medical image compression/decompression using vivado HLS (High level synthesis). Data contrast compression method has been proposed to perform image compression using vivado HLS, HLX design blocks for implementation using Arty z7 kit. Data contrast compression method provides efficient image compression by performing ‘C/C++’ code in vivado HLS environment and effective design blocks are designed using VHDL code in vivado HLX environment. Performances of above two environments are implemented in Arty z7 20 kit. Results for Software simulation, Hardware simulation, design blocks, synthesis, implementation, bit generation, elaborated design or schematic view are generated accordingly. Hardware implementation result can be viewed in Arty z7 kit. Performance parameters like compression ratio, latency are concentrated and tabulated accordingly. Medical Image compression applications are widely used in the field of telehealth for storage and communication purpose.

Analysis of Factors Influencing Bagging Performance using Supervised Machine Learning Method

Ari Primantara, Udi Subakti Ciptomulyono, Berlian Al Kindhi — 2024-03-26

Urea fertilizer is an essential product mainly for people working in an agriculture field. The demand increases every year caused by rapidly growth globalization. This problem obliges fertilizer industry to be more competitive by reducing products losses and production cost. This study introduces supervised machine learning to predict which factors influence fertilizer weight losses during bagging process, such as moisture, products temperature, and wind pressure. Method used in this study is Random Forest Regressor, a well-known method that combines several decision trees into a single output commonly for classification and regression. Factors studied were analysed using Random Forest resulting MSE, R² and Spearman’s method to determine the correlation between the factors towards fertilizer weight. MSE value obtained from this study was 0.0036 and R² -0.5334. Low R² result may be caused by insufficient data. The Spearman coefficients of moisture, products temperature, and wind pressure were 0.035, 0.244, 0.013, respectively. Spearman coefficient shows good result if the value ranging from -1 to +1, which obtained from products temperature. This study shows that Random Forest Regressor can predict several factors that influence productivity of fertilizer industry, particularly in bagging system. However, further research with larger data range is still needed

Gold Rush Optimization-Driven Random Forest Approach for Intrusion Detection System in Edge and Fog Computing Settings

2024-03-26

Although both edge and fog computing architectures improve latency minimization and real-time data processing, their scattered nature presents significant security problems, particularly in intrusion detection. In this study, we provided an innovative gold rush optimization-driven random forest (GRO-RF) technique for effective intrusion detection systems (IDS). To examine the performance, the UNSW-NB15 public dataset is utilized to train the suggested GRO-RF technique. The z-score normalization approach is used to preprocess the raw samples to rearrange the data without noise and duplicates. To extract important features, the cleaned data is subjected to additional processing throughout the feature extraction process using linear discriminate analysis (LDA). The RF technique is used to identify intrusions in edge and fog computing environments using the retrieved data. GRO is designed to enhance the misclassification of RF by increasing accuracy and reducing the error rate in categorization. The suggested method is implemented using a Python program and its efficacy in detecting intrusions is evaluated against other current approaches using various metrics such as precision (98.45%), accuracy (99%), F-measure (96.89%) and recall (97.25%). We show that, in edge and fog computing scenarios, the GRO-RF technique has the highest intrusion detection accuracy compared to the other methods, based on the results of the experiment.

Trustworthy Intra Cluster Management Scheme (TICMS) to Improve Lifetime of Wireless Sensor Networks

S. Suresh Babu, N. Geethanjali — 2024-03-26

Wireless Sensor Networks (WSN) place a significant emphasis on energy efficiency because power consumption is the single most important factor in determining the overall lifespan of the network. There have been multiple suggestions made for potential strategies that could reduce the energy consumption of nodes. In this article, a intra cluster management scheme based on trust is provided and it is demonstrated that the method is also energy efficient. The success of the work is dependent on significant phases, which are the network area segregation phase, Cluster building phase includes cluster head node selection, computation of the trust degree & the control of the node's state and the routing phase. Taking into account the location coordinates allow for the formation of the cluster and the CH node selection. The CH node is responsible for calculating the trust level and maintaining control over the state of the other nodes in the network. The network lifetime is significantly improved, owing to the control over working nodes. The experimental results show that the proposed work is satisfactory with regard to the longevity of the network and energy efficiency.

Node Clustering and Cluster Head Selection in Multi-Channel Wireless Sensor Networks

Siddhartha Adhyapok, Bhaskar Bhuyan, Anand Sarma, Hiren Kumar Deva Sarma — 2024-03-26

Hierarchical approach for overall management of a multi-channel wireless sensor network seems to be more practical in terms of stability, scalability and also reliability. Clustering of nodes in such a hierarchical network plays an important role. Moreover, appropriate selection of cluster head nodes can not only improve the network performance but it can result in prolonged lifetime of the network too. In this paper, a node clustering algorithm for multi-channel wireless sensor network is proposed along with a methodology for selection of the respective cluster head nodes. The node clustering algorithm may be executed at sink and nodes are clustered considering two different parameters namely geographic proximity and availability of common channels. The nodes inside a cluster are expected to be geographically close to each other and also they are desired to have access to maximum number of common communication channels. Access to common communication channels shall reduce the overhead due to channel switching. Again at the time of selection of the cluster head nodes, the principles of Analytic Hierarchy Process (AHP) are exploited and as per AHP principles, the most suitable node is selected as the cluster head for a cluster of nodes. The performance of the proposed protocol has been evaluated and compared with few benchmark techniques available in literature. The proposed approach outperforms other protocols in terms of energy efficiency, throughput, end-to-end delay, communication overhead, network lifetime, and re-clustering time. The future scopes of the work are outlined.

An Impact of Deep Learning and Machine Learning Technologies in Image Processing

2024-03-26

In recent years, the integration of deep learning and machine learning technologies has revolutionized the field of image processing, leading to significant advancements in various applications. From object detection and recognition to medical image analysis and remote sensing, these technologies have reshaped how images are processed, interpreted, and utilized. This paper explores the profound impact of deep learning and machine learning in image processing, highlighting key advancements, challenges, and future directions.

Integrating Features and Unlabeled Data with Modified Support Vector Machines for Improved Lung Cancer Detection

Suman Antony Lasrado, G. N. K. Suresh Babu — 2024-03-26

This research explores the application of Modified Support Vector Machines (MSVMs) as a potent classifier for the effective diagnosis of lung cancer, aiming to enhance the accuracy and performance compared to conventional Support Vector Machines (SVMs). While SVMs have been widely employed, their limitation lies in treating all features equally, potentially affecting the precision of disease detection. In response to this, MSVMs introduce a novel approach by incorporating both labeled and unlabeled data into the learning process, gradually searching for the optimal separating hyper plane. The key innovation lies in the assignment of weights to a kernel function, measuring the importance of individual features and addressing the shortcomings of traditional SVMs. By acknowledging the varying significance of features, MSVMs offer a more explored and efficient classification process. The newly formulated kernel function enables the integration of labeled and unlabeled data, contributing to a more robust learning model. The proposed modification not only enhances the classifier's ability to discern between malignant and benign lung tissues but also opens avenues for improved pattern recognition indicative of lung cancer. The research investigates the comparative performance of MSVMs against different SVMs, with preliminary results indicating promising outcomes. The integration of both labeled and unlabeled data, combined with the consideration of feature importance through weighted kernel functions, demonstrates the potential of MSVMs as a breakthrough in the accurate classification of lung cancer. While further validation with larger datasets is essential, this study suggests that MSVMs could emerge as a significant advancement in the field of lung cancer diagnosis, offering heightened 93% accuracy and 99% specificity in predicting and classifying the lung cancer disease.

A Comprehensive Review of the Artificial Neural Networks (ANN) Methodology Implementations for Analysing and Forecasting the Efficiency of Solar Water Heater Collector Under Various Tilt Angles

Chaitali S. Deore, Sagarkumar J. Aswar — 2024-03-26

This article examines tilt angle and reviews artificial neural network (ANN) models for tilt angle (TA) prediction and optimization in solar water heaters (SWH). The paper provides a summary of design simulations, parameters, applications, and mathematical approaches that are used in a variety of applications. The quantity of references analysing TA deployment in context of research publications are increasing. The number of countries involved in solar system operations has increased dramatically. Many models and test procedures for determining the optimal TA in various solar schemes has been created, each of which is identified by their mathematical models and tracking techniques, as evidenced by recent research. The 4 ANN models like Extreme learning machine (ELM), radial basic function (RBF), Multilayer Feed forward neural network (MLFNN), and Artificial Neuro-fuzzy inference System (ANFIS) are compared by estimating the root mean square error (RMSE) value of training and testing of models. The ELM performs better compared to other models. The variables of TA like slips, inclination, height, and width are also mentioned in this article. The analysis of SWH at various tilt angles and development of ANN model for optimization of the TA is also discussed in this article. Here 35 research paper related to development and analysis at various tilt angle has been reviewed and understand their impact of different ANN approaches on the performance of SWH.

Intrusion Detection System by Integrating Mod K- Means C Algorithm and T-SNE Dimensionality Reduction

Mallaradhya C., G. N. K. Suresh Babu — 2024-03-26

This research presents a new methodology to enhance the precision of Intrusion Detection Systems (IDS) by integrating the Modified K-Means Clustering (ModKMeansC) algorithm as a classifier along with t-Distributed Stochastic Neighbor Embedding (t-SNE) a reduction technique of dimensionality. In the realm of cyber security, conventional IDS face challenges in accurately discerning abnormal network behavior due to the dynamic and intricate nature of cyber threats. The ModKMeansC algorithm, intricately designed to address issues stemming from abnormal network connections, introduces parallelism into centroid and distance calculation update operations. This concurrent execution, performed asynchronously for each data point, facilitates real-time analysis of network traffic, thereby bolstering efficiency and responsiveness. Leveraging the CICIDS2017 dataset, encompassing both normal and abnormal network traffic patterns, the study implements and fine-tunes the ModKMeansC algorithm for optimal performance. t-SNE is applied to preprocess the data before feeding it into the classifier. The proposed system's performance is meticulously assessed using key performance metrics. A proportional analysis against traditional intrusion detection algorithms underscores the ModKMeansC algorithm's advantages in accurately categorizing abnormal network behavior as 92%. Results and ensuing discussions highlight the algorithm's adeptness in efficiently handling abnormal network connections and its prowess in parallel processing. This examination significantly supports to the dynamic field of cyber security by presenting a more effective and responsive methodology for identifying abnormal network behavior. The amalgamation of the ModKMeansC algorithm with t-SNE holds considerable promise in elevating the accuracy of IDS as 95%. Future research directions may encompass adapting the proposed system to real-world cyber security scenarios and further optimizing the algorithm for scalability in large-scale networks.

Artificial Intelligence and Machine Learning to Enhance E-Money Utilisation and Human Resource Development in Ica Agro-Export Firms

2024-03-26

The integration of Artificial Intelligence (AI) and Machine Learning (ML) technologies has revolutionized various industries, and their potential impact on e-money utilization and human resource development is substantial. This paper explores the synergies between AI/ML and e-money systems, elucidating their role in optimizing financial transactions and bolstering human resource capabilities. In the realm of e-money, AI algorithms can analyze vast datasets to discern patterns, detect anomalies, and predict consumer behavior. By leveraging these insights, e-money platforms can offer personalized financial services, streamline transactions, and mitigate fraud risks. Furthermore, AI-powered chatbots and virtual assistants enhance user experience by providing real-time support and personalized recommendations, fostering greater trust and engagement within e-money ecosystems.Moreover, AI and ML technologies hold immense promise for human resource development. Through data-driven insights, organizations can optimize talent acquisition, identify skill gaps, and tailor training programs to individual needs. AI-driven recruitment platforms streamline the hiring process by automating candidate screening and matching, thereby expediting the identification of top talent. Additionally, ML algorithms facilitate continuous learning initiatives by analyzing employee performance data to deliver personalized learning experiences and targeted skill development pathways..

Machine Learning and Artificial Intelligence for the Development of Social Responsibility and Risk Management Techniques

2024-03-26

In an era marked by unprecedented technological advancements, Machine Learning (ML) and Artificial Intelligence (AI) are emerging as powerful tools for promoting social responsibility and enhancing risk management practices. This paper explores the transformative potential of ML and AI in addressing societal challenges and fortifying risk mitigation strategies. The intersection of ML and AI with social responsibility endeavors opens avenues for proactive engagement and impactful interventions. Through sentiment analysis and social media monitoring, AI algorithms enable organizations to gauge public perceptions, identify emerging issues, and tailor their initiatives to address societal needs effectively. Moreover, ML-powered predictive analytics facilitate data-driven decision-making, enabling businesses to anticipate and respond to social and environmental risks proactively. Furthermore, AI and ML technologies offer novel approaches to risk management across various domains. In the financial sector, predictive modeling and algorithmic trading algorithms enhance risk assessment and portfolio optimization, bolstering resilience against market fluctuations. In healthcare, ML algorithms analyze patient data to identify potential health risks and optimize treatment strategies, thereby improving patient outcomes and reducing healthcare costs. However, the adoption of ML and AI for social responsibility and risk management also poses ethical and regulatory challenges.

Application of Artificial Intelligence in Education: The Role of Technology as an Educational Tool

Samsul Susilawati, Triyo Supriyatno, Sutiah — 2024-03-26

Because of its potential to completely transform conventional teaching methods, artificial intelligence (AI) has attracted a lot of attention from a variety of industries, including education. This essay addresses AI's potential as a teaching tool and examines its use in education. The main emphasis is on how AI technology may improve overall educational outcomes, personalise education, and enhance learning experiences. The paper starts off by giving a general introduction to artificial intelligence (AI) and its main ideas, such as deep learning, machine learning, and natural language processing. After that, it explores particular uses of AI in education, like chatbots that can tutor students and platforms that adapt to their needs. These technologies make use of AI algorithms to evaluate student data, offer individualised feedback, and design personalised learning pathways based on each learner's requirements and preferred method of learning. The study also looks at the advantages and difficulties of incorporating AI into educational environments. Benefits include increased access to a wealth of instructional resources, better learning efficiency, and more student involvement. To guarantee ethical and fair AI use in education, issues including algorithmic bias, data privacy concerns, and the requirement for teacher training in AI utilisation must also be addressed. The article also explores possible developments and future trends in AI-driven education, including AI-powered educational assistants, personalised learning ecosystems, and virtual reality simulations. These developments have the power to completely change education by improving accessibility, effectiveness, and engagement for students of all ages and backgrounds..

Machine learning Evaluation on Effects of Transformational Judgement and Performance metrics in Information Industry

Wiliam, Syaifuddin, Sofiyan, Salman Faris — 2024-03-26

In the rapidly evolving landscape of the information industry, understanding the impact of transformational judgment on performance metrics is crucial for organizational success. This study employs machine learning techniques to evaluate the effects of transformational judgment on various performance metrics within the information industry. Transformational judgment, defined as the ability to envision and enact transformative strategies, is examined as a predictor variable, while performance metrics such as efficiency, innovation, and customer satisfaction serve as outcome variables.Using a dataset encompassing a diverse range of information industry organizations, this study applies regression and classification algorithms to analyze the relationships between transformational judgment and performance metrics. Feature selection and engineering techniques are employed to enhance model accuracy and interpretability. Additionally, model evaluation metrics such as accuracy, precision, recall, and F1-score are utilized to assess the predictive performance of the machine learning models. This research contributes to both theoretical understanding and practical applications within the information industry by elucidating the importance of transformational judgment in achieving organizational success. By leveraging machine learning techniques for predictive analysis, organizations can identify and cultivate transformational leadership qualities to optimize performance outcomes and gain a competitive edge in the dynamic information landscape.

Artificial Intelligence based Emotional Intelligence for data Analytics

Indra Kesuma, Zainuddin, Sofiyan, Salman Faris — 2024-03-26

In recent years, the integration of Artificial Intelligence (AI) and Emotional Intelligence (EI) has emerged as a promising avenue for enhancing data analytics. Emotional Intelligence, a vital human trait involving the recognition, understanding, and regulation of emotions, offers a unique dimension to AI-driven analytics by infusing machines with empathetic capabilities. This paper explores the convergence of AI and EI in the context of data analytics, elucidating how incorporating emotional understanding into AI systems can revolutionize data interpretation and decision-making processes. The primary focus of this paper is to delineate the potential benefits and challenges associated with leveraging emotional intelligence in data analytics through AI algorithms. By harnessing EI, AI systems can better comprehend human emotions expressed in textual data, social media interactions, and other unstructured sources, thereby providing deeper insights into consumer sentiment, market trends, and user behavior. Furthermore, AI-driven emotional intelligence can enhance personalized recommendations, improve customer service interactions, and facilitate more empathetic human-machine interactions.

Artificial Intelligence for English Learning Enhancing Vocabulary Acquisition

2024-03-26

The acquisition of vocabulary is a fundamental aspect of language learning, particularly in English, which serves as a lingua franca in global communication. Traditional methods of vocabulary acquisition often rely on rote memorization and repetition, which can be tedious and ineffective for many learners. With the advancements in artificial intelligence (AI) technologies, there exists an opportunity to revolutionize vocabulary acquisition through innovative and personalized approaches. This paper explores the potential of AI in enhancing vocabulary acquisition for English learners. It begins by examining the shortcomings of traditional methods and the challenges faced by learners in vocabulary acquisition. Subsequently, it delves into the various ways in which AI can be leveraged to address these challenges effectively. One of the primary advantages of AI in vocabulary acquisition is its ability to provide personalized learning experiences tailored to individual learners' needs and preferences. Through adaptive algorithms and machine learning techniques, AI platforms can analyze learners' strengths, weaknesses, and learning styles to deliver customized vocabulary exercises and content. This personalized approach not only increases engagement but also maximizes retention and understanding.

Machine Learning Based Toolbox in Foreign Language for Children to Address Climate Change Adaptation

2024-03-26

Climate change poses a significant threat to our planet, and educating future generations about its implications and solutions is paramount for effective adaptation and mitigation efforts. However, language barriers can hinder the dissemination of crucial information, particularly to children who may not yet be proficient in the predominant language of scientific discourse. This paper proposes a novel approach to addressing this challenge by developing a machine learning-based toolbox in a foreign language tailored for children. Leveraging advances in natural language processing and educational technology, the toolbox aims to facilitate interactive learning experiences in foreign languages, fostering a deeper understanding of climate change and promoting actionable strategies for adaptation. Machine learning algorithms like k-nearest neighbor, decision tree, logistic regression, and deep learning techniques such as natural language processing and artificial neural networks are being utilized to tackle climate change challenges across different sectors, including transportation

Deep Learning based Task Prediction and Neural Network Analytics for Employees

Vidya Dwi Amalia Zati, Syaifuddin, Sofiyan, Salam Faris — 2024-03-26

In today's dynamic work environments, accurately predicting tasks and optimizing employee performance are crucial for organizational success. Traditional methods often fall short in handling the complexity and variability of modern workplaces. This paper proposes a deep learning-based approach to task prediction and neural network analytics for enhancing employee productivity and efficiency.Our methodology leverages deep learning architectures such as recurrent neural networks (RNNs), convolutional neural networks (CNNs), and transformers to model the intricate relationships between various factors influencing task assignments and employee performance. By integrating diverse data sources including historical task assignments, employee profiles, project requirements, and performance metrics, our model learns complex patterns and dependencies, enabling accurate task predictions and insightful analytics.Key components of our approach include data preprocessing to handle noise and missing values, feature engineering to extract relevant information, and model training using large-scale datasets. We explore techniques such as attention mechanisms to capture salient features and interpret model predictions. Additionally, we employ transfer learning to leverage pre-trained models and adapt them to specific organizational contexts, facilitating faster convergence and improved performance.

Examining the Influence of Artificial Intelligence between Burnout, Employee Performance in Higher Education

Maria Linda, Zainuddin, Erikson Saragih, Salman Faris — 2024-03-26

Higher education is just one of several sectors that have seen a meteoric rise in the AI integration rate in recent years. Within the setting of academic institutions, this literature review seeks to investigate the effects of artificial intelligence on burnout and productivity among staff members. Academic burnout, which impacts health and work satisfaction, is on the rise and is defined by emotional tiredness, depersonalization, and diminished personal achievement. However, educational institutions and the quality of education they provide are highly dependent on the performance of their employees. In the first part of the study, we take a look at the big picture of artificial intelligence (AI) in higher education and how it may improve administrative duties, learning experiences (via personalized suggestions), and decision-making. There are worries that workers will be displaced from their jobs, have their responsibilities altered, and have more work to complete as a result of AI's fast adoption.

Machine Learning Based Genetic Algorithm to Design Job Rotation Schedules Ensuring Homogeneity in Industry 4.0

Hendry, Syaifuddin, Sofiyan — 2024-03-26

The advent of Industry 4.0 has ushered in an era of dynamic workforce management, necessitating innovative approaches to optimize human resource utilization while maintaining a harmonious work environment. Job rotation, a practice aimed at diversifying employees' tasks, has emerged as a crucial strategy to enhance skill development and mitigate workplace monotony. This paper presents a novel framework leveraging machine learning-based genetic algorithms to design job rotation schedules that ensure homogeneity across various dimensions within the workforce. The methodology begins with the identification of key parameters, including skill sets, experience levels, and ergonomic considerations, to construct a comprehensive representation of the workforce landscape. Subsequently, a genetic algorithm, guided by machine learning models, iteratively generates and refines job rotation schedules to minimize disparities in workload distribution, skill utilization, and fatigue levels among employees. Through iterative optimization, the proposed framework strives to achieve a balance between organizational objectives, such as productivity enhancement and employee satisfaction, while adhering to operational constraints. The efficacy of the approach is demonstrated through simulation studies and real-world case analyses, highlighting its potential to facilitate agile workforce management in the context of Industry 4.0. Overall, the integration of machine learning and genetic algorithms offers a promising avenue for designing job rotation schedules that promote workforce homogeneity, resilience, and adaptability in the dynamic landscape of modern industries.

Enhanced Security Encryption and Data Driven Model for Digital transition using Artificial Intelligence

Dang Thanh Le, Nguyen Van Thanh — 2024-03-26

In an era of digital data security, ensuring robust security measures is paramount to safeguarding sensitive data. This paper proposes an innovative approach combining enhanced security encryption and a data-driven model empowered by artificial intelligence (AI) to fortify digital transitions. The methodology begins with a meticulous assessment of objectives, data inventory, and classification to discern the scope and sensitivity of information involved. An encryption strategy tailored to the data's sensitivity level is then devised, encompassing encryption at rest, in transit, and end-to-end encryption where applicable. Integrating AI into the security framework enables real-time threat detection through sophisticated algorithms analyzing network traffic, user behavior, and system logs. Moreover, AI-driven behavioral analytics augment monitoring capabilities, enabling the identification of anomalies indicative of potential security breaches. By amalgamating encryption with AI-driven insights, this approach presents a comprehensive solution to fortify digital transitions, ensuring data integrity, confidentiality, and compliance with regulatory standards.

Artificial Intelligence and Machine learning using Classification Method for Building models

Bakhrani Abdul Rauf, Mithen Lullulangi, Onesimus Sampebua, Rahmansah — 2024-03-26

Artificial Intelligence (AI) and Machine Learning (ML) techniques, particularly classification methods, have gained significant traction across various domains for building predictive models. Classification algorithms are essential components of AI and ML systems that enable the categorization of data into predefined classes or labels based on their features. This paper provides an overview of AI and ML methodologies, focusing specifically on the utilization of classification methods for constructing robust and accurate predictive models.The primary objective of this paper is to elucidate the principles and applications of classification techniques in AI and ML. It explores popular classification algorithms such as Decision Trees, Support Vector Machines (SVM), Logistic Regression, Random Forest, and Neural Networks, detailing their underlying mechanisms, advantages, and limitations. Furthermore, the paper discusses the preprocessing steps, feature engineering techniques, and model evaluation methods associated with classification-based model development.Through real-world case studies and examples, this paper demonstrates the versatility and effectiveness of classification algorithms in solving diverse problems, including image recognition, text classification, sentiment analysis, medical diagnosis, fraud detection, and customer churn prediction. It highlights the importance of data quality, model interpretability, and domain knowledge in the successful implementation of classification-based AI and ML solutions.

Retracted

Retracted — 2024-03-26

Retracted

A Transfer Learning Approach for Bipolar Disease Detection

R. Saranya, S. Niraimathi — 2024-03-26

Bipolar illness is a complicated mental health issue that affects a large section of the world. Effective bipolar illness therapy requires early and precise diagnosis. This work proposes a unique transfer learning strategy for bipolar disorder identification using many modules to improve prediction accuracy. The first module trains a CNN, BiLSTM, and RBF. Deep learning architectures help this module extract relevant characteristics from raw input data. The second module uses a DNN for feature selection to enhance feature representation. The DNN model eliminates superfluous or duplicated data by identifying the most important bipolar disorder diagnosis characteristics. In the third module, transfer learning uses a pre-trained model. Pre-trained models improve bipolar illness prediction by using learnt representations. Transfer learning is modified to include domain knowledge from related activities or datasets. The fourth module implements a RESNET Classification module. RESNET excels in picture categorization; therefore we use it to forecast bipolar disorder by capturing complicated data patterns and correlations. In the fifth module, SGD optimizes the model. By repeatedly adjusting parameters based on a portion of training data, SGD speeds convergence and improves accuracy. Finally, we optimize Levy Flight-based Fruit fly optimization to fine-tune model parameters. This technique optimizes hyper parameters including learning rate, batch size, and regularization for optimal bipolar illness identification.

Evaluation of Diabetic Medications Using Hybrid Fuzzy Pattern Classifier and TOPSIS

Soren Atmakuri Davidsen, M. Padmavathamma — 2024-03-26

Given the plethora of pharmaceuticals available to regulate blood glucose levels, in medical decision-making, choose which ones to take for Type 2 Diabetes is a difficult task. Making decisions is made more difficult by the variety of hyperglycemia-lowering medications, each of which has distinct benefits and potential drawbacks. The study proposes a fuzzy Multi-Criteria Decision-Making model-based computer-aided healthcare decision-making system. This methodology combines the full multiplicative form of the TOPSIS method with Ratio Analysis and a modified version of Fuzzy Multi-Objective Optimization. The goal is to help with the decision-making process while choosing Type 2 Diabetes pharmaceutical therapy. The Fuzzy TOPSIS approach analyzes each option by taking into account all criteria in compliance with a published clinical guideline, On the other hand, while applying the TOPSIS technique to determine the relative relevance of particular criteria, expert opinions are taken into account. In order to address the drawbacks of conventional reference points and improve the ranking process in fuzzy multi-criteria decision-making, this study investigates an extended reference point technique inside the hybrid MCDM paradigm. The principal medicine, DPP-4-I, is confirmed by computational results, and Metformin is recognized as the second-line add-on therapy. The third, fourth, and fifth options are sulfonylurea, glucagon-like peptide1 receptor agonist, and insulin, in that order. To assess the effectiveness of the model, a sensitivity study is carried out by contrasting the outcomes with previous research, different fuzzy MCDM approaches, and an interval TOPSIS method based on observational data. Endocrinologists should be aware of the substantial agreement found between the final anti-diabetic drug rankings produced by the proposed hybrid model and alternative approaches.

SFA-BCPF: Selective Forwarding Attacks in WSN With Bayesian Confidence-Based Packet Forwarding Algorithm

K. Soundarraj — 2024-03-26

Wireless Sensor Networks (WSNs) play a pivotal role in various applications, ranging from environmental monitoring to military surveillance. However, their vulnerability to malicious attacks, particularly Selective Forwarding Attacks (SFAs), poses a significant challenge to the reliability and integrity of the transmitted data. SFAs involve compromised sensor nodes selectively dropping or delaying certain packets, leading to the degradation of network performance and undermining the overall efficiency of WSNs. This research proposes a novel approach to counteract SFAs by introducing a Bayesian Confidence-Based Packet Forwarding Algorithm (BCPF). The algorithm leverages Bayesian probability theory to dynamically assess the trustworthiness of each sensor node in the network. By considering factors such as historical behavior, communication patterns, and data integrity, the algorithm assigns a confidence level to each node. Nodes with higher confidence levels are prioritized for packet forwarding, while those with lower confidence levels are subjected to additional scrutiny or avoided altogether. The Bayesian Confidence-Based Packet Forwarding Algorithm aims to enhance the robustness of WSNs against SFAs by promoting the forwarding of packets through nodes with proven reliability. This research contributes to the ongoing efforts to fortify WSNs against emerging security threats, fostering their continued deployment in critical applications.

Internet of Things based Type 2 Diabetes Prediction using Enhanced Feed Forward Neural Network with Particle Swarm Optimization

S. Arulananda Jothi, J. Abdul Samath — 2024-03-26

The Internet of Things (IoT) is an emerging network that enables everyday objects to connect to the web and exchange and collect data. The IoT is crucial in healthcare because it allows for constant patient monitoring and informed decision making. Diabetic complications now impact a sizable fraction of the population. The elderly are disproportionately affected by type 2 diabetes, which is also the most prevalent form of the illness and which is associated with a wide range of serious health issues such as cardiovascular disease, renal failure, blindness, stroke, and even death. That's why knowing the patient's prognosis or receiving a diagnosis quickly may help. Improving the prediction model's accuracy takes time and work, but one of the biggest challenges is figuring out how to properly analyze the data to get the right conclusion. Many models may be employed for analysis; for instance, many Neural Network models have been used for clinical diagnosis. The problem is that these models haven't improved much, in terms of either accuracy or precision, whether in the training or testing stages of sickness diagnosis. This study offers an Enhanced Feed forwarded Neural Network (EFNN) that employs a chaotic-based particle swarm optimization model (EFNNCPSO) to analyze IoT-based datasets. The proposed method has the potential to improve the accuracy of predicting Type 2 diabetes in an IoT environment. The suggested network is able to learn all of the features in the dataset and performs efficient calculations. Finally, analogous models to the one proposed are compared. The proposed EFNNCPSO has a higher accuracy than state-of-the-art methods (99.9%).

Effect of Non-Newtonian Convective Unsteady MHD Flow of Fluids on an Inclined Plate in a Porous Medium with Boundary Slip and Heat Source

Rajni, Monika Kalra — 2024-03-26

In this research paper I have calculated and computed some of the important aspects of Impact of Convective, Unsteady MHD Flow of Fluids having Non-Newtonian Character, on Inclined Plate in a Medium of Porous Nature, with Boundary Slip, Including Heat Source. The theme of this research paper is to investigate non- Newtonian character on porous material plate in inclined position with the slip condition. The motive of investigation of this research paper is to find out certain rare results for the fluid dynamics. The problem has been solved analytically using mat lab by solving various equations culminating through process. Velocity, temperature, skin friction values/ profiles have been obtained in the wake of non-dimensional parameters such as magnetic, permeability including heat source along with Prandtl and Grashof number The present problem deals with non-Newtonian unsteady, electrically conducting fluid MHD, convective flow, on Semi-infinite, inclined plate of porous material, held in oscillating position in porous medium, with suction, over which the fluid is allowed to flow while a transverse magnetic field acted on it and velocity slip condition applied on it at the boundary. It will help the researchers for future research and experiments with different non-Newtonian fluids to find impact of slip flow using plate in inclined position.

ULODF: An Unsupervised Learning based Outlier Detection Framework in High Dimensional Data

C. Jayaramulu, Bondu Venkateswarlu — 2024-03-26

Outliers play crucial role in applications like disease diagnosis, fraud detection techniques and cyber security to mention few. Unsupervised learning techniques like clustering are widely used, in the area of machine learning, towards outlier detection. However, most of the existing methods did not consider dual tasking benefits of using clustering that not only renders quality clusters but also identifies outliers effectively. We proposed a framework named Unsupervised Learning based Outlier Detection Framework (UL-ODF). An algorithm named Novel Outlier Detection Method in High Dimensional Data (NODM-HDD) is defined. The algorithm has mechanisms to improve compactness of clusters made besides determining outliers. The algorithm exploits an enhanced version of K-Means clustering technique. A prototype is built to validate the utility of the framework and the underlying algorithm. Different benchmark datasets and metrics are used in the empirical study. The experimental results revealed that the NODM-HDD shows better performance over the state of the art.

Smart City Infrastructure Monitoring using AI and IoT Technologies

Vaishali V. Sarbhukan (Bodade), Jyoti S. More, Yogesh Jadhav — 2024-03-26

Smart cities are developing to effectively manage resources, improve citizen services, and assure sustainable development as civilizations rapidly urbanize. Advanced infrastructure monitoring technologies like AI and IoT are essential for smart city development. This article proposes an AI and IoT-based smart city infrastructure monitoring design. The framework includes IoT sensors on highways, bridges, buildings, water supply systems, and electricity networks throughout the city. These sensors record structure health, traffic movement, ambient conditions, energy usage, and other characteristics in real time. A centralized AI-powered monitoring system analyzes and decides on this data. The monitoring system's AI algorithms handle massive sensor data using machine learning, deep learning, and predictive analytics. The system can detect structure degradation, transportation congestion, and environmental dangers in real time using pattern recognition and anomaly detection. Predictive analytics lets you schedule maintenance and optimize resource allocation. The monitoring system's intelligent decision support helps municipal managers make infrastructure management and emergency response choices. Critical occurrences trigger automated warnings and messages to avoid accidents or minimize interruptions. The suggested framework increases municipal infrastructure operating efficiency and public quality of life by assuring safety, dependability, and sustainability. Smart cities can create a more connected, resilient, and sustainable future by using AI and IoT.

Texture Volume of Fractions Using Integration System

Deepali Saxena — 2024-03-26

A fundamental tool in the domains of material characterisation and image processing, capture analysis makes it easier to extract important information from pictures and data representations. The emergence of three-dimensional (3D) imaging technology has led to the demand for reliable quantitative metrics for precisely characterizing the textural characteristics of volumetric objects. Texture Volume of Fractions (TVF) is one such metric that offers a potent way to quantify the arrangement and dispersion of textures in three-dimensional objects. In order to give a thorough grasp of the idea, mathematical expression, and importance of TVF in texture analysis, this work focuses on investigating the usage of integration systems for TVF computation. As a quantitative metric for 3D texture characterization, TVF is introduced. TVF provides insights into the composition and spatial distribution of textures inside 3D objects, in contrast to standard approaches that only use two-dimensional representations. Researchers and practitioners may better grasp how texture attributes are quantified and expressed in the context of 3D image analysis by establishing the mathematical formulation of TVF.

Then, as a unique strategy to precisely and efficiently computing TVF, the integration system approach is introduced. Integration systems take use of advances in computer approaches to make texture analysis more efficient, allowing researchers to work with vast amounts of data in an efficient manner. The benefits of integration systems—such as increased computational effectiveness and noise resistance—are thoroughly explored, emphasizing how they may improve texture characterisation in a range of 3D image processing applications.

Efficient Hybrid Load Balancer for Software Defined Networks using OpenFlow Accuracy Prediction

Ananth B. — 2024-03-26

Cloud computing is a global vision for real-world IT offerings where data and resources are integrated by web-based cloud management organizations using hardware and structured, primarily web-based packages. people at a reasonable cost. Sharing resources can cause problems with access to those resources, leading to a crash. The strategy for distributing network traffic across multiple connecting node or servers is called as load balancing. It is referred that no server is overloaded. Load control builds user responsiveness by distributing shares evenly. It also makes projects and sites more accessible to customers. The reason for this archive is to understand the billing control. It has associated structures of communication organizations over the Internet. Load balancing is an important part of a distributed computer to stay away from work overload and provide equally important support. Different statistics are used to determine system complexity

Impact of Different Data Management Frameworks on Common Data Management Tasks in Information System (R Language Perspective)

Anant Prakash Awasthi, Niraj Kumar Singh, Masood H. Siddiqui, Aanchal A. Awasthi — 2024-03-26

To maximize data processing and analysis, effective data management is essential. It ensures that data is efficiently processed, readily accessible, secure, and well-organized. This enhances data integrity, reduces the amount of redundancy, and it makes decision-making more prompt. In an era where data is a valued asset that drives innovation and strategic decision-making, effective data management techniques are essential.

The two essential data management activities for improving data processing are joining and sorting. By combining datasets based on common characteristics, joining makes thorough analysis easier. Sorting data well enhances search and retrieval. When combined, these processes enhance the accuracy and speed of data processing, simplifying workflows and enabling sound decision-making. Database management systems depend on joining and sorting to enable the creation of value, the extraction of significant insights, and the identification of trends from massive datasets.

The performance of native R, tidyverse, and data.table when merging data in R varies. Large datasets may cause Native R to lag, despite its versatility. Known for its readability, Tidyverse strikes a balance between performance and simplicity. Because of its exceptional speed, Data.table is a very effective option for large-scale data joins. The decision is based on the complexity and amount of the dataset. The best option for maximum performance, particularly for complex and large-scale jobs, is Data.table. Native R and Tidyverse work well with smaller, more manageable datasets when code readability is crucial. Every method addresses particular requirements in R data analysis. Similarly, when it comes to sorting data in R, Native R, tidyverse, and data.table behave differently. While Native R provides a standard method, it might not be as effective with larger datasets. Although readability is given priority in Tidyverse's user-friendly syntax, it may not be as fast as more efficient options. Once more, Data.table runs faster and uses less memory when sorting large amounts of data than the competition. The decision is based on the needs of the analysis: data.table for best performance, especially with large datasets and computationally intensive tasks; tidyverse for readability; and Native R for simplicity.

Hence, in order to sum up, effective data management is essential for businesses to fully utilize their data and make wise decisions. Optimizing data processing and analysis requires careful consideration of joining, sorting, and tool selection.

Data Privacy and Compliance Issues in Cloud Computing: Legal and Regulatory Perspectives

Satyanarayan Kanungo — 2024-04-13

Cloud computing has revolutionized how organizations store, process and share data. However, the use of cloud services introduces complex data privacy and compliance challenges from legal and regulatory standpoints. This paper explores the key data protection laws and regulations impacting cloud computing, including the EU's General Data Protection Regulation (GDPR), the California Consumer Privacy Act (CCPA), the Health Insurance Portability and Accountability Act (HIPAA), and industry-specific requirements like the Payment Card Industry Data Security Standard (PCI DSS). It examines the shared responsibility model between cloud providers and customers, jurisdictional considerations, international data transfers, vendor management, incident response obligations, and auditing/monitoring of cloud environments. The paper also discusses evolving trends such as the increased focus on data localization laws and the growing adoption of secure enclaves and confidential computing. Finally, it provides recommendations for organizations to navigate this complex landscape through robust governance frameworks, risk assessments, contractual safeguards with cloud service providers, and transparency with end-users. Effectively addressing data privacy and compliance issues is essential for organizations to reap the benefits of cloud computing while protecting sensitive information and upholding their legal and ethical duties.

Machine Learning Techniques Based on Ensemble Feature Selection for Disease Indentification and Classification in Plant Leaves

Aniruddhsinh Dodiya — 2024-03-26

Farmers have a number of challenges when trying to examine vast regions for plant diseases manually. This is because it takes a lot of time and needs a big number of experienced labourers with a true grasp of plant diseases. In order to diagnose plant diseases accurately and quickly, image processing and machine learning models might be used. Agricultural specialists now use visual or microscopic examination of leaves or certain chemical methods to diagnose plant diseases. Large farms need a large crew of specialists and continual plant monitoring, both of which are prohibitively costly for the typical farmer. Managing plant diseases is essential for increasing crop yields and ensuring a healthy food supply. To begin with, GCMO, or Graph Cut-based Multi-level Otsu, is a variation of unsupervised multi-stage segmentation that this study suggests. It combines Graph Cut and Multi-Level Otsu algorithms. After that, several evaluation metrics are used to compare the segmentation performance of the proposed technique with current unsupervised segmentation algorithms. The pictures of rice, peanut, apple, and potato plant leaves are used for this purpose. When compared to preexisting unsupervised methods, the segmentation accuracies achieved by the suggested approach were much higher when evaluated on a variety of conventional and real-time datasets. This study's primary goals are to apply deep learning-based classification to pre-processed results and to adopt optimum Feature Selection (FS) to segmented ones. Kernel Fuzzy C Means (KFCM) is used for leaf image segmentation and the multi-level Otsu Thresholding approach is used for impacted area segmentation after the input pictures are pre-processed using a multiscale retinex algorithm.

A Statistical and Machine Learning Based Face Identification System with Enhanced Multiple Weighted Facial Attribute Sets

Shreyas Patel, — 2024-03-26

Academic and business institutions alike have been more interested in facial recognition studies in recent years. The idea of face recognition has grown in significance in several applications due to its openness and the myriad of security characteristics it encompasses. Face recognition solves several issues with alignment, age, lighting, emotion, and lighting. The aforementioned challenge arose while trying to differentiate one face from another in a facial recognition system. This study proposes a novel method for improving face recognition performance using Multiple Weighted Facial Attribute Sets in conjunction with the Principal Component Analysis (PCA) methodology. The results of this study demonstrate that the recognition system's overall performance was affected by the weights assigned to the various qualities. During the matching process, the user-defined input component of the proposed approach will prioritise a collection of picture characteristics.

Do Gen Z buy Cosmetics using Augmented Reality Impulsively? A Deep Learning-Based SEM-ANN Analysis

Anish Kaushal — 2024-03-26

Social media marketing is effective tools in relation to Generation Z’s impulse purchasing behaviour within fashion industry in the context of cosmetics purchase. However, there is research gap on mediating roles of media types and ease of payments on social platforms in awakening the interest of Gen Z females and moderating roles of virtual reality (VR). SEM-ANN was conducted based on 287 valid responses. The findings reveal that media can trigger a stronger urge to buy impulsively and impulse buying intention. Moreover, the awakening of interest and ease of payment options plays mediating roles in this process. Secondly, with the recent advancements of virtual reality (VR) technologies, VR applications in cosmetics are increasingly becoming an inevitable trend of future shopping. The empirical results show that media types and ease of payments results in creating urge to impulsive purchase behaviour. Furthermore, the role of AR and VR strengthens the impulsiveness trait in driving the urge to buy impulsively.

Efficient Machine Learning-Based Drowsiness Detection for Enhanced Driving Safety: Real-Time Implementation

Pradeep Laxkar — 2024-03-26

In today’s rapid time changing era, the count of road accidents is increasing day by day because of sleeping disorders and drowsiness. Technical enhancement in each and every area of day-to-day life, also demands the technically enhanced driving cars which detect drowsiness in driver with more accuracy and efficiency. This study presents a real-time drowsiness detection system for drivers, by blending the power of machine learning techniques to analyze facial features like Pupil of eye, EAR, MAR and NLR, considering the system (Car) watch, GPS system as well as utilizing the Advanced Driver Assistance Systems (ADAS) of smart cars. The system employs OpenCV and Dlib to extract eye, mouth aspect ratios and nose length ratio from video frames with the other gained feature of smart cars. The data undergoes standard scaling preprocessing before training a deep neural network for binary classification of drowsy and non-drowsy states. The model architecture comprises four dense layers with dropout and L2 regularization, ending in a softmax activation. Stratified K-Fold cross-validation is utilized for data splitting, and the model is compiled using the Adam optimizer and categorical cross-entropy loss, incorporating an early stopping callback to mitigate overfitting. The proposed system demonstrates exceptional performance, achieving more than 99% accuracy, 0.993 recall, and 0.991 F1 score in real-time drowsiness detection. These results hold potential for enhancing road safety and reducing fatigue-related accidents by accurately identifying drowsiness in drivers. With a capacity to detect drowsiness in real-time at a level of high accuracy, the proposed system has an immense potential to increase road safety and prevent accidents related to fatigue.

Efficient Hybrid Load Balancer for Software Defined Networks using OpenFlow Accuracy Prediction

Ananth B. — 2024-03-26

Inception-v3 vs. DenseNet for Automated Detection of Diabetic Retinopathy

Tajender Malik — 2024-03-26

The purpose of this paper is to explore the effectiveness of automated detection methods in diagnosing diabetic retinopathy (DR), a leading cause of vision loss among individuals with diabetes. By leveraging advancements in artificial intelligence and image processing techniques, the study aims to assess the accuracy and efficiency of automated systems in identifying retinopathy, thus enabling early intervention and improved patient outcomes. A comprehensive review of existing literature on automated detection systems for DR was conducted. Various image analysis algorithms, including deep learning approaches and feature extraction techniques, were explored and evaluated based on their performance in detecting retinal abnormalities associated with DR. In this research, we present an Inception-v3 and DenseNet-based automated detection technique for DR using retinal fundus pictures. This work involves the training, evaluation, and comparison of the performance of DenseNet and Inception-v3 convolutional neural networks (CNN) on a publicly available dataset of retinal fundus images. Inception-v3-based classifiers have performed better than DenseNet-based classifiers with the same dataset. While DenseNet achieved classifier accuracy and precision of 89.2% and 89.6%, respectively, Inception-v3 has been able to achieve classifier accuracy of 95.8% and precision of 95.9%. Inception-v3 has also exceeded area under ROC in comparison to DenseNet by 0.3% in two categories. The findings of this study highlight the promising potential of automated detection methods for DR. The integration of automated systems in clinical settings has the potential to enhance early diagnosis, facilitate timely treatment interventions, and improve patient outcomes.

Statistical Approach in Indian Capital Market through Quantitative Modeling of Quarterly Financial Metrics Using Deep Learning

Ashish Garg — 2024-03-26

The use of computing power in stock market analysis has been a popular field of research for investors and retailers seeking to maximize their profits from the market. However, there is limited research on the relationship between quarterly financial results and future stock price movements for companies listed on the National Stock Exchange in India. This study aims to fill this knowledge gap by examining this relationship and analyzing the impact of key technical and fundamental parameters on future stock prices. Data scraping techniques were used to collect quarterly results and stock price data, and the analysis showed that the proposed model provided an average profit of 142% over a three-year period, with an annual profit of 34.7%. The neural network model achieved a 62.9% accuracy on the test dataset. Improvement opportunities exist for higher accuracy. The experimental results demonstrate that the proposed model can play a vital role in stock price prediction and could be useful for investment decision-making. Overall, this study provides valuable insights into the impact of stock fundamentals on stock prices and could be a valuable resource for investors and retailers seeking to maximize their profits from the stock market.

An Expert System for Diagnostic Construction Problems in Pavement (Flexible, Rigid, Composite , Block or Brick Pavement) in Addition to Subgrade and Subbase Condition Defects (ES-DATPs)

Muhammad Khidre Musa, Shahban Ismael Albrka — 2024-03-26

Construction of roadway pavement entail significant issues that are impacted by various factors and are nearly impossible to overcome without expert assistance. Treatment and maintenance such challenges in pavement and providing the best economical solutions requires substantial technical competence, which owing to scarce resources and far-flung locations might not be accessible at every construction site. Establishing an expert system in this scope is a highly impressive method for assisting debutant engineers in overcoming and learning about these challenges. Interviews and questionnaires are used to obtain more expert information. This information is recorded, analyzed, represented, and converted to computer application through the Visual Studio language of programming, and the system is valled as ES- DATPS. The main aim of creating this expert system (ES- DATPS) is to helps highway engineers and civil engineers to acquaint causes, prevention, degree of severity, testing, measurement, maintenance and repair of all problems for (rigid pavement, flexible pavement, composite pavement, block or brick pavement) in addition to subbase and subgrade defects with accuracy and detail. According to the degree of severity, the best practice and economic method for maintenance each type of pavement can be selected by engineers. This expert system helps highway engineers to manage challenges and recognize type of defects by figures and video

The Influence of Different Factors on the Iraqi Construction Industry Using Organizational Culture: An Intelligent Measurement Model Assessments in AMOS-Based SEM

Alhamza Yassin Flaih Maeni — 2024-03-26

The positive impact of the construction industry's efficiency on cost reduction, economic growth, and overall national economies is a widely recognized phenomenon across the globe. The implementation of construction projects is often impeded by various limitations and hazards that impede the progress of project operations, ultimately resulting in a significant adverse impact on the overall performance of the projects. Prior studies have assessed the impact of different factors pertaining to a company or project while disregarding the present study aims to address the research gap by incorporating organizational culture as a moderating variable and assessing the impact of different factors, namely stakeholders, communication, cost, technology, top management support, and local authority support, on the performance of the construction industry in Iraq. The research data was collected through a survey questionnaire with several construction practitioners in Iraq. The AMOS Software is used to analyze the data and develop the measurement model. A survey instrument was utilized to obtain information for the research from a multitude of construction companies in Iraq. The data have been subjected to analysis, and a measurement model has been constructed using AMOS 26 to test the results of the hypotheses. The results reveal that the model fits the nature of the data and the research variables. Investigating the effect of these factors will help the construction industry to prevent or mitigate risks, control expenditures, and achieve competitive advantages. Measuring the Effects of Different Factors on the Iraqi Construction Industry Using Organizational Culture.

Predicting Mathematics Incompetence Effects on the Study of Digital Electronics Among Electrical and Electronic Engineering Students, using Artificial Neural Networks

Theodore Oduro-Okyireh — 2024-03-26

Through proficiency effect analysis, the research aims to identify key engineering mathematics domains that are essential for students to succeed in Digital Electronics course. This investigation employs an artificial neural network (ANN)-based predictive model and focuses on Ghanaian Technical Universities as a case study. The study adopted the quantitative research design where random cluster sampling was used to select a total of 488 final year Higher National Diploma students from four technical universities in Ghana. The data consisted of mathematics achievement test scores and results of their Digital Electronics course. After testing a number of artificial neural network (ANN) architectures, the most accurate model was a multilayer perceptron neural network (MLPNN) with a 6-2-2 partition. The results showed, with high precision, that Functions and Algebra are two critical areas of mathematics that have the greatest impact on students’ performance in Digital Electronics in electrical and electronic engineering studies.

Feature Edge Extraction for Human Body Cut Injured Detection using Deep Learning

Chandrashekhar Kumar — 2024-03-26

In recent endeavours, the management style of medical patient treatment has been growing continuously. In the meantime, trends demand more accuracy and work performance beyond the level of advancement for waiting in line in hospitals, ending with emergency cases as well. To overcome the delay in response from the hospital end with medical high demand, the doctor-patient ratio is also around 1:700. And the need for blood in a human case is a little crucial based on the blood group. Also, to avoid this scenario and save human lives, we have to implement blood detection on the spot using the image processing technique with deep learning, which can close the gap between the arrangement of the blood bank system and human blood safety support with the patient as well. The current study can only be done after reaching the hospital; using various research attempts to fill this gap. But delay is still needed to reach proper support with advanced technology. To extract the image of blood colour detection using an optical clamp-on Sensor(OCS) deep learning blood detection and amount of blood flow based on healthy body range of blood status with a range of 5 litters for the adult body with advanced BMI technology.

Segmentation in Cervical Cancer Detection: A Key Step in Early Diagnosis

Pothineni Syam Sowbhagya Sree — 2024-03-26

Cervical cancer refers to a type of cancer that develops in the cells of the cervix, which's the lower part of the uterus connecting to the vagina. Many cancers affect people all over the world. One of them is cervical cancer. Preventing the disease requires early detection and successful treatment rather than recognizing the issue at an advanced stage. These precautions can help prevent deadly cancer and contribute to a healthy life. This cancer can be treated well if it is detected early by a medical checkup for HPV lesions and risk factors for malignant cervix formation It is commonly triggered by the papillomavirus (HPV) a sexually transmitted infection. Globally cervical cancer ranks as the most prevalent cancer among women with around 570,000 new cases being diagnosed every year. Fortunately, this form of cancer is highly preventable through screenings and HPV vaccinations effectively reducing the risk of its development. Our research paper primarily focuses on enhancing cancer diagnosis and analysis by employing various techniques such as Contour segmentation, fitness score assessment, detection rate calculation, identification of optimal threshold values, geometric mean analysis, ROI examination, and three-segnet architecture. According to our research, we achieved a detection rate of 85%, a fitness score of 95%, a geometric mean of 90%, and positive results in the ROI examination. As a result of improving our techniques, we can provide better results for all images, resulting in better diagnosis and treatment. Continuing to innovate in medical imaging is crucial for providing the best possible care for cervical cancer patients.

Machine Learning Model for Optical Character Recognition-Based Food Allergen Detection with Recommendation System for Alternative Food

Rugved Borade, — 2024-03-26

In today’s diverse and fast-paced food industry, ensuring consumer safety and meeting specific dietary needs is of paramount importance. Food allergen detection and recommendation systems have emerged as crucial tools to address these concerns. This project aims to create an innovative OCR based solution for automating the identification of allergenic ingredients on food packaging labels. By combining Optical Character Recognition (OCR) technology with a comprehensive allergen database, the system will provide real-time allergen information to consumers. Moreover, it will recommend suitable food alternatives for individuals with specific dietary restrictions, enhancing their shopping experience and reducing the risk of allergen-related incidents. The allergen knowledge base is implemented using several machine learning algorithms and will be updated constantly. This holistic approach not only promotes food safety but also empowers consumers to make informed choices, fostering a healthier and more inclusive food environment.

Unlocking the Potential of Quantum Machine Learning: A Paradigm Shift in Optimization

Saurabh Choudhary, — 2024-03-26

Quantum Machine Learning (QML) is an exciting new field that combines quantum computing and machine learning, revolutionizing the way we develop systems. This article explores the significant role QML plays in traditional communication and quantum optimization methods. We delve into the fundamentals of quantum computing, compare classical methods with quantum optimization, and examine QML algorithms to illustrate their applications across different industries.

¹ Krishna Engineering College, Ghaziabad, India; email – saurabh.chy75@gmail.com

³ Krishna Engineering College, Ghaziabad, India; email – rajneesh8m@gmail.com

² Krishna Engineering College, Ghaziabad, India; email – sachin.malhotra2312@gmail.c

⁴ Athenaeum Jupiter Pvt Ltd, Gurugram, India; email – nitin@athenaeducation.co.in

⁵ Lincoln University College, Malaysia; email – midhun.research@gmail.com

*Correspondence: Sachin Malhotra

HoD, Department of Computer Science & Engineering Krishna Engineering College, Mohan Nagar,Ghaziabad, India

Zip Code - 201007

Email: sachin.malhtra2312@gmail.com

Phone: +91 9911217804

With a solid understanding of quantum mechanics and machine learning concepts, our research breaks down optimization techniques, highlighting their advantages and disadvantages compared to quantum methods. We introduce QML algorithms, such as quantum neural networks and quantum approximate optimization algorithms, and provide explanations of their workings.

Moving beyond theory, we demonstrate how QML can effectively address real-world optimization problems in finance, transportation, healthcare, and other domains. Our examples showcase how QML can enhance performance, reduce costs, and foster innovation. Despite its potential, the integration of QML into daily business faces challenges. We explore issues such as hardware limitations, error correction, scalability, and noise reduction. Additionally, we present potential solutions and suggest future research directions to overcome these challenges.

In summary, our research underscores that QML, as a fusion of classical and quantum optimization, is poised to transform business practices and drive innovation. As quantum hardware advances and our understanding of quantum algorithms deepens, the game-changing capabilities of QML will revolutionize our approach to complex development problems, propelling progress and innovation across various industries.

Detecting and Eliminating Blackhole Attacks for Improving the Quality of Service of Mobile Ad-Hoc Networks Using RTS-CTS Mechanism

Ganesh Dhondu Dangat — 2024-03-26

Mobile Adhoc Network is one of the infrastructures-less, decentralized wireless networks that can reconfigure by itself. MANET does not depend on the access points in the network, where it can accommodate any existing infrastructure. Since it is an ad-hoc network, all the nodes in the MANET are mobile nodes connected wirelessly. Depending on the routing protocols, single-hop, two-hop, and multi-hop-based data transmission is followed in the MANET. These models provide more opportunities for malicious activity creation in the network, where it destroys data transmission and loss. Several earlier research works have focused on detecting and eliminating malicious activities. One of the dangerous attacks that defy detection of the physical properties is the Blockhole attack, and they don't respond to its sender and receiver nodes. Compared to other malicious attacks, blackhole attacks result in a small amount of data loss in the network, and they are considered a major research problem in MANET. This paper has aimed to provide a better solution using the RTS-CTS mechanism and initialize the data transmission with dummy data to detect the black hole nodes. Once the black nodes are identified in the network, they are eliminated immediately, and their functionalities are with the neighbor nodes. The simulation results obtained from NS2 show that the proposed RTS-CTS mechanism outperforms and provides better QoS.

Influential Nodes Identification in Complex Networks: Sampling Approach

Karzan K. Abdulmajeed — 2024-03-26

Accurately identifying influential nodes within complex networks is crucial for understanding information and influence propagation. Existing state-of-the-art algorithms, while powerful, often rank all nodes, which can be computationally expensive and unnecessary for many applications. In this paper, we propose a simple yet efficient approach that overcomes these limitations. Initially, a systematic sampling methodology was employed to strategically select a subset of nodes from the network, representing a small fraction of its entirety. Subsequently, the betweenness centrality of these sampled nodes was estimated to facilitate their ranking. To assess the performance of our sampling method alongside alternative algorithms, we employ the stochastic Susceptible–Infected–Recovered (SIR) information diffusion model to compute various metrics including the infection scale, the final infected scale over time, and the average distance between spreaders. Our experimental findings, conducted on real-world networks, indicate that our proposed method accurately identifies influential nodes while maintaining significant computational efficiency.

Mitigating Generic Attacks for Intrusion Detetction System Based on CGAN and FIPSO Using UNSW-NB 15 Dataset

Vineeta Srivastava, — 2024-03-26

: In recent times, there has been a notable surge in cyberattacks due to the Internet of Things' exponential growth. Because of this, maintaining corporate borders today requires cybersecurity. Intrusion detection systems, or IDSs, are used to notify users of noteworthy events when maintaining a network. The first is the identification of malicious traffic, for which zero-day attack detection research is essential. This research provides an improved intrusion detection model that leverages FIPSO for feature extraction, conditional Generative Adversarial Networks (cGAN) to handle data imbalance, and machine learning techniques for classification tasks. We evaluated the model for binary and multi-classification, focusing on the UNSW-NB15 dataset in particular. The proposed methodology is noteworthy because it employs Random Forest (RF) classification along with FIPSO to enhance feature selection and cGAN to directly address the issue of data imbalance. This hybrid technique yields better results, with 83% accuracy in multi-class classification and 96% accuracy in binary classification.

Enhancing Network Security through Machine Learning-Based Anomaly Detection Systems

Salam Allawi Hussein — 2024-03-26

For decades, anomaly detection has been used to discover and extract aberrant components from data. Several techniques have been employed to spot irregularities. Machine learning (ML) is a method that is gaining importance due to its significant significance in this area. Machine learning models that detect anomalies in their application are the focus of this study's Systematic Literature Review (SLR). In our investigation, we look at the models from four angles: how anomaly detection is classified, what it's used for, how machine learning is done, and how well machine learning models perform. In this study, we looked for papers published in 2015–2023, which deal with the topic of anomaly detection using machine learning techniques. After we've finished analyzing the selected research papers, we'll go on to outline 10 different uses of anomaly detection that were found in those publications. The number of machine learning models used to detect anomalies is also identified, accounting for 6% of all instances. Finally, we offer available a wide range of datasets used in anomaly detection studies as well as many other generic datasets. Furthermore, compared to other categorized anomaly detection methods, researchers are more likely to employ unsupervised anomaly detection. The application of machine learning models for anomaly detection is one of the most promising fields of study, and researchers have utilized several ML models in this regard. Therefore, based on the results of this review, we advise and suggest things to researchers.

CdO Doped CuO for Gas Sensing Environmental Applications

Mohammad Arqam — 2024-03-26

This paper describes the hydrothermal synthesis of cadmium/copper oxide nanocomposite (CdO-CuO). XRD, FESEM, and EDS were used to determine the structure, morphology and compositional. Furthermore, the LPG sensing capabilities of pure and CdO doped CuO were investigated. The sensors had sensitivity of about 24 and 36 at 275oC operating temperature and 250 ppm gas concentration. LPG is extremely flammable even at very low ppm concentrations; detecting LPG low ppm concentrations is needed for security reasons.

Robustness Analysis for hyGWO-PS Optimized FOPID-Controllers in AGC of Interconnected Hydro-Thermal Power System

Shailaja Yogesh Kanawade — 2024-03-26

It has already been found in the literature that the hybrid grey wolf optimization- pattern search (hyGWO-PS) tuned fractional order PID (FOPID)-controllers in three area interconnected hydro thermal power system (TAIHTPS) with nonlinearities, multiple tie lines and reheat turbines has produced the far better performance than some recent published approaches. In that study, the settling times and overshoots of frequency & tie line power deviations and ITAE values were obtained by proposed approach called hyGWO-PS/FOPID under the nominal condition and are evaluated as: Settling time of ∆f₁ = 8.50s; Settling time of ∆f₂= 8.50s; Settling time of ∆f₃= 8.10s; Settling time of ∆P_Tie12 = 19.31s; Settling time of ∆P_Tie23= 15.23s; Settling time of ∆P_Tie31=13.01s; ITAE=1.1243. In this regard, it has become necessary to study the variation in the performance of TAIHTPS consisting of hyGWO-PS optimized FOPID-controllers with parametric variations, i.e. with varying load conditions and system parameters (T_G, T_T, T_R, T_Wand T₁₂). In the present work, the robustness analysis or the sensitivity analysis of hyGWO-PS optimized FOPID-controllers under parametric variations for AGC of same interconnected power system has been carried out. The robustness analysis shows that the behaviour or the system dynamic responses of TAIHTPS consisting of hyGWO-PS optimized FOPID-controllers hardly alters under the variations in operating load conditions and system parameters over the range [-50%, +50%], i.e. hyGWO-PS optimized FOPID is far better robust for the same.

Driving Style Recognition for Intelligent Vehicle using Unsupervised Clustering Algorithms

Abhishek Dixit — 2024-03-26

The manner in which a driver operates their vehicles has a significant impact on both energy management and driving safety. Moreover, it is a crucial factor in the advancement of driver assistance systems (ADAS), which aim to increase the level of vehicle automation. As a result, numerous research and development initiatives have been undertaken to identify and classify driving styles. In this study, we have used principal component analysis for feature reduction and K means clustering algorithm for driving style identification of the vehicle. To evaluate the performance of the proposed approach, it was tested using vehicle trajectory data from the Next Generation Simulation (NGSIM) project, specifically the datasets collected on US Highway 101 and I-80. The proposed approach introduces a novel method that enhances efficiency and accuracy, offering a significant advancement in addressing complex challenges within its respective domain.

Harnessing Artificial Intelligence for Effective Corporate Governance: Evaluating the Board of Directors Role and Its Influence on Individual Investors and ESG Practices

Swathi G. — 2024-03-26

As the landscape of corporate governance undergoes a transformative shift in the digital age, this research paper investigates the integration of Artificial Intelligence (AI) tools into governance frameworks. Focusing on the pivotal role of the Board of Directors, the study delves into the multifaceted impact of AI on corporate decision-making processes. Concurrently, it explores the intricate relationship between corporate governance, individual investors, and Environmental, Social, and Governance (ESG) practices. The primary objectives are to assess the evolving responsibilities of the Board, analyze the deployment of AI tools in governance structures, and understand their collective influence on the awareness levels of individual investors regarding ESG considerations. The literature review navigates through traditional and contemporary corporate governance models, elucidating the role of the Board of Directors and the dynamic integration of AI. Drawing upon theoretical frameworks such as agency theory, stakeholder theory, and behavioral finance, the paper establishes a conceptual foundation for the ensuing empirical investigation. Methodologically, the research employs a comprehensive approach, encompassing sample selection, data collection methods, and analytical techniques to scrutinize the intricate interplay of AI, boards, and investors. The subsequent sections delineate the historical evolution of the Board's responsibilities, shedding light on the challenges and opportunities presented in the digital era. Simultaneously, an in-depth analysis of AI applications in corporate governance, including case studies and ethical considerations, provides a nuanced perspective. Individual investors' perceptions of ESG factors are scrutinized, emphasizing the impact of corporate governance on investor trust. The research concludes with a synthesized exploration of the interdependencies between AI, the Board of Directors, and individual investors, offering insights into the transformative potential of AI in enhancing corporate governance practices. The study contributes to the existing literature by unraveling the challenges, risks, and future implications associated with this paradigm shift. Recommendations for future research and practical implications provide a roadmap for stakeholders navigating the dynamic intersection of AI, corporate governance, and investor relations.

Enhancing Healthcare Image Security with DNA Cryptography in the IOT Environment

Animesh Kairi — 2024-03-26

The integration of DNA cryptography into IoT-based healthcare image systems represents a groundbreaking approach to tackle the pressing issue of data security and patient privacy in the digital healthcare landscape. With the increasing reliance on the Internet of Things in healthcare, the need for robust security measures is paramount to protect sensitive medical information. This research explores the feasibility and efficacy of employing DNA cryptography to encode medical data, adding an additional layer of security to data transmitted over IoT networks while ensuring compatibility with existing healthcare infrastructure. By enhancing data integrity, particularly in the realm of E-healthcare, this innovative cryptosystem has the potential to facilitate secure patient data transfer between healthcare institutions, ultimately advancing patient care and preserving privacy. The promising integration of DNA cryptography underscores its potential as a secure and rapid technique, offering a hopeful solution to the security challenges faced by healthcare image systems within IoT environments.

An Efficient and Optimized Backoff Scheme for Disseminating Heterogeneous Traffic for Multiple Instances of LLN-Based Industrial IOT Networks.

Animesh Giri — 2024-03-26

With numerous devices being connected to the Internet to create a network of smart things, the Internet of Things (IoT) has experienced rapid growth in recent years. However, because of their constrained power, processing, and bandwidth, these devices frequently require assistance with network efficiency and dependability.
In order to overcome the difficulties of distributing heterogeneous traffic across numerous instances of low power and Lossy-IoT networks, this research suggests an effective, optimized backoff scheme. The suggested method makes use of backoff algorithms in conjunction with network coding to increase network efficiency and reliability while decreasing transmission delay.
The binary exponential-backoff (BEB) algorithm and the truncated binary exponential backoff (TBEB) algorithm are two of the backoff algorithms used in the suggested scheme. The BEB algorithm is used to resolve collisions during transmission, while the TBEB algorithm is used to reduce the backoff stage and the transmission delay.
The suggested scheme also makes use of network coding, which raises the network’s dependability by enabling multiple nodes to work collectively to transmit data, reducing the likelihood of data loss, and guaranteeing that the data reaches its target location.
Extensive simulations are used to evaluate the performance of the proposed scheme, and the results show that it performs better than traditional backoff algorithms in terms of network efficiency, reliability, and transmission delay. The outcomes also show that the suggested method can manage heterogeneous traffic, which qualifies it for IoT networks with a variety of devices and applications. In conclusion, the suggested effective, optimized backoff scheme offers a potentially viable answer to the problems faced by lossy, low-power IoT networks. The scheme’s combination of backoff algorithms and network coding enhances network efficiency and reliability while reducing transmission delay, making it an attractive option for IoT network deployments.

Comparative Analysis of Various Textual-Visual Models for Self-Attentive Query Focused Video Summarization

Sheetal Girase — 2024-03-26

The exponential growth of video data presents a significant challenge in extracting pertinent information from it. Video summarization aims to address this issue by extracting essential information from video data in order to facilitate the exploration of videos. Given the subjective nature of determining "relevant information" in a video based on user preferences, it is imperative to establish a mechanism that takes into account the users' preferences during the process of generating a summary. One approach that can be employed is to enable users to input a query. Rather than generating a predetermined and inflexible summary for a given video input, this study has investigated a method of generating a video summary that caters to the preferences of the user. Query Focused Video Summarization (QFVS) is regarded as a supervised learning problem in the context of the YouTube Dataset [4]. It aims to produce a summary based on user inputs, specifically the video and the textual query. The query relevance of frames from the video is determined by mapping them to a shared multimodal semantic embedding space. By utilising our attention network and encoder, we have successfully enhanced the accuracy rate from 61.91% [4] to 74.60%. Extensive experiments were conducted utilising deep learning models, specifically ResNet34 and DenseNet, to extract image features. Additionally, word2vec and GloVe were employed for word mappings. The integration of textual and image features is employed for diverse experimental purposes.

“Application of Lavaan for Structural Equation and Path Model on Unmet Healthcare Needs Latent from Demographic and Health Survey Data of Maharashtra State, India”

Kishor N Raut — 2024-03-26

Structural Equation Modeling (SEM) is useful in many areas of research including Medical Research. In R Project, the package lavaan have commercial-quality with fully free open-source to use for latent variable analysis. This study aimed to analyze the strength of association latent variable (V623 (Exposure), V624 (Unmet need) and S253 (Had your uterus removed) with other endogenous latent variable of Demographic and Health Survey Program 2019-21 Phase VII, data of Maharashtra State, India. The women age ranged from 16 to 49 years.

Path model of endogenous latent variable f7 (V623 (Exposure), V624 (Unmet need) and S253 (Had your uterus removed)) measured the effect of f1, f2, f3, f4, f5 and f6 latent variable in addition to V613 and V625A exogenous variable, The study revealed that for latent model f7 needs each latent was built-up on sufficient (3 or more) variables in each latent to avoid the Heywood case, even after study data having sufficiently enough large of observation data.

Artificial Intelligence Systems in Managing Human Resources: An Exploratory Study in the Indian Context

Shambhavi Pandey — 2024-03-26

This paper is an attempt to examine the factors affecting the use of Artificial Intelligence (AI) by human resource professionals with their ‘professional experience (in years)’ as a moderating variable. A survey research conducted on a sample of 123 senior human resource professionals. The key findings reveal that the use of AI would lead to a lack of employee productivity, morale and trust. It further illustrates that AI would have adverse consequences on growing employee silence and data manipulations. HR practitioners in India in are differing in adopting AI is not because of their fear of losing their jobs but because of the sheer nature of unpredictable outcomes and lack of strong legislations on using AI. This study answers the question of why there is not a widespread use of artificial intelligence systems in India for managing human resources even though AI is being used for other domains of management.

Analysing the Characteristics of MIMO Antennas: Enhancing Isolation and Employing Soft Computing Techniques via a Systematic Review

Nagarjuna Tanikonda — 2024-03-26

MIMO systems play a crucial role in communication systems due to their reliance on antenna design for the air interface. As network evaluations continue, there have been advancements in network capacity and technology to meet user requirements. Recent literature focusing on the MIMO system has revealed that metamaterials and soft computing approaches offer improved performance for mobile communication. This review encompasses the literature on the MIMO system, employing various design models and covering standard journals from 2016 to 2023. All research studies discussed in this review utilized tools such as HFSS, CST, and MATLAB. Additionally, a concise examination is conducted on isolation enhancement and soft computing approaches. Furthermore, a comparative analysis is provided for diversity measures including isolation, gain, axial ratio bandwidth, ECC, and CCL. The challenges and future directions highlighted in this review serve as inspiration for antenna researchers to develop MIMO antennas tailored for mobile communication.

A Novel Non-Dominated Sorting Dragonfly Optimization With Evolutionary Population Dynamics Based Multi-Objective Approach For Feature Selection Problems

Anitha G — 2024-03-26

Feature selection is a multi-objective problem which includes two contradictory objectives. It is an effective method in classification to eradicate noise, inappropriate and redundant features to maximize the classification precision and reduce the number of chosen features. In this study, meta-heuristic algorithm with multi-objective approach have been tried to explore feature selection problem with a combination of non-dominated sorting dragonfly algorithm and evolutionary population dynamics strategy. First, to enhance the value of non-dominated solutions, an evolutionary population dynamic strategy is integrated with a heuristic natural selection operators. Second, to avoid the local optimum trap and enrich the population variety, to upgrade the step size and to maintain exploration and exploitation balance, a strategy is planned to optimize these issues. Finally a Pareto optimal solutions are obtained from the non-dominated sorting strategy which makes the algorithm appropriate for handling multi-objective feature selection problems. Simulations are performed on 18 datasets from UCI repository. The proposed NDSDA, NDSDA_EPD and NDSDA_EPD_CM approaches are compared with the existing dragonfly algorithms. The proposed algorithms outperforms the other techniques by enhancing the grouping accuracy and decreasing the preferred features count.

Deep Learning-Based Classification Methods for Detection of Diseases in Rice Leaves – A Review

Prameetha Pai — 2024-03-26

Cultivating rice is crucial in India to meet demands of a growing population. In order to improve crop yield, it's essential to address factors like diseases caused by bacteria, fungi, and viruses. Detecting and managing these diseases is vital, and one effective approach is employing rice plant disease detection methods. Deep learning techniques, known for their ability to analyse data, are used for disease identification in plants. This work explores various deep learning approaches for detecting rice plant disease. Deep learning, particularly in computer vision, has shown significant progress in detecting plant diseases. The study compares the effectiveness deep learning mechanisms, demonstrating superior performance of deep learning models. Utilizing deep learning can help prevent major crop losses by detecting leaf diseases through image analysis.

Optimizing Corrosion Prediction Initiation Time for Embedded Steel in Concrete with Shell Powders Using Deep Learning Techniques

Lavanya M R. — 2024-03-26

The embedded steel is integrated with concrete material, primarily used in buildings and infrastructure projects. "Embedded steel" refers to steel reinforcement bars or mesh embedded in concrete structures. Steel is added to the concrete to strengthen and support the structure. One of the primary challenges associated with embedded-based steel is anticipating its corrosion once it has been incorporated into building structures. It is necessary to monitor the initiation time of corrosion on the steel in the concrete, which is considered crucial to the environment. Early corrosion detection is challenging, and its accuracy helps design durable concrete. This process reduces the time and cost of embedded steel manufacturing. This research focuses on applying embedded deep-learning models to test the accuracy of the algorithms suggested for embedded steel. A-state of art technique reveals that convolutional neural network (CNN), Long short-term memory (LSTM), and Deep neural network (DNN) models can perform accurate predictions. In this study, the above deep learning models are embedded to validate the accuracy of the different algorithms. The study aimed to determine the corrosion initiation time on steel, which is Incorporated within concrete via corrosion potential measurement. To achieve this, concrete samples were arranged with conch shell powder as a partial replacement to Portland cement and exposed in 5% sodium chloride with following the requirements of ASTM C876 – 15. During the exposure time, the steel embedded's corrosion potential was measured, and the resulting dataset was utilized for training three deep-learning models. These models were developed using input variables such as cement, conch shell powder, fine aggregate, coarse aggregate, exposure period, and water to estimate the corrosion initiation time on the embedded- steel based on the potential corrosion measurements.

Yakshagana Character Identification Through Deep Learning with Crown and Facial Makeup Pattern Analysis

Anantha Murthy — 2024-03-26

Yakshagana, an intricate theatrical art form originating from Karnataka, encompasses variations such as Thenku-thittu, BadaguThittu, and Badaabadagu Thittu. This research delves into the historical roots, contemporary influence, and evolving makeup trends within Yakshagana. Within the Tenkutittu Yakshagana, diverse crown types take center stage, with the performer's chosen crown and facial makeup pattern serving as key determinants of the portrayed character. Our study focuses on character classes including Vishnu, Devi, Sarpa and Mahisha for character identification. To address the intricate task of character classification in Yakshagana images, this paper employs deep learning methods such as Three Tier CNN and YOLOv5 . Specifically, a Cyclic Gate Recurrent Neural Network is utilized to classify characters like Vishnu, Devi, Sarpa and Mahisha. Following character categorization, the model proceeds to determine disguises. The Three-tier CNN achieves a commendable 90% accuracy in classifying disguises. Through thorough testing, it has been established that YOLOv5, boasting a remarkable 95% accuracy in identifying multiple elements within an image, emerges as the most suitable algorithm for character identification. This research serves as a real-time tool, aiding newcomers in identifying the appropriate crown and makeup pattern for specific Yakshagana figures.

Fuzzy Logic Based Grid Integration of Photovoltaic/Wind Hybrid Power Generation

Srikanth D — 2024-03-26

The study outlines a hybrid solar-wind system that uses three-phase power grid architecture to ensure sustainable and effective power generation. The hybrid solar-wind device uses the one of the technique of Maximum Power Point Tracking (MPPT) to optimize total effectiveness at the Common Coupling Point (PCC) by combining a photovoltaic station with a wind farm. This guarantees the best possible energy production with wind and solar power systems, regardless of the weather. The 3-phase neutral point clamped multilevel inverter's DC-link voltage is adjusted using a fuzzy logic controller that has been designed and proven to follow the vector control technique. This ensures that the inverter maintains the intended level. The hybrid system MATLAB/SIMULINK is used to execute the simulation, comparing the performance of the PI controller and the FLC controller overall. Step reaction of the MPPT and DC-link voltage technique performance are included in the assessment. The findings demonstrate that, in spite of fluctuating weather conditions, the FLC controller effectively maintains a grid voltage, achieves a power factor of one, and makes the most of the use of the solar-wind hybrid energy system injection by employing active power. In conclusion, the research presents a thorough method to maximize power production and improve overall performance of a hybrid solar-wind system. It makes use of the FLC controller and MPPT technique to manage the DC-link voltage in an efficient manner for the best performance in various weather conditions. The design of effective renewable energy integration of renewable energy sources and systems into the electrical grid are both greatly enhanced by the suggested approach.

An Innovative Approach for Revolutionizing Pediatric Health Monitoring in Real-Time Activity Recognition Utilizing CNN-LSTM-ELM

Preethi Salian K — 2024-03-26

Pediatric activity recognition is an essential part of many healthcare and childcare applications, allowing for the monitoring and evaluation of children's physical development. In this study, a novel real-time pediatric activity recognition system is proposed, which combines the advantages of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks for extraction of features, followed by an Extreme Learning Machine (ELM) classifier for accurate activity categorization. It initially generated an extensive dataset made up of footage of kid-friendly activities that had been carefully labelled with activity categories. A two-step procedure is employed, starting with the use of a CNN model to extract discriminative spatial features from video frames that has been pre-trained on a large dataset. The image signals available in pediatric activities are richly represented by these elements. In order to capture temporal relationships within the series of feature vectors, it incorporates an LSTM network after feature extraction. Further improving the recognition accuracy, this LSTM-based sequence modelling is skilled at identifying subtle activity patterns and transitions over time. The key component of this development is the addition of an ELM classifier after the LSTM layer. ELM, which is renowned for its ability to train quickly and effectively, utilizes the temporal context stored by the LSTM to conduct real-time activity classification with astounding speed and accuracy. As a result, pediatric actions are recognized effectively and robustly. The CNN-LSTM-ELM model is utilized to analyze receiving images in order to do real-time recognition. The system is equipped with this framework to enable real-time decision-making in scenarios including healthcare and child care. The findings show that the suggested CNN-LSTM-ELM architecture demonstrates outstanding accuracy of 90.5% and efficiency in identifying a wide spectrum of pediatric activities, hence enhancing the capabilities of child-focused healthcare and wellbeing applications