Adaptive SVM with Bio-inspired Optimization Tuning for Guava Leaf Disease Prediction

Abstract

In recent years, precision agriculture has increasingly adopted intelligent systems to monitor plant health and detect diseases at an early stage.  Guava a widely cultivated tropical fruit, is highly susceptible to leaf diseases such as Anthracnose, Rust, and Pestalotiopsis. These diseases not only reduce crop yield but also degrade fruit quality, directly affecting farmers' income. Traditional disease detection methods rely heavily on manual inspection, which can be time-consuming, subjective, and ineffective at scale. Consequently, there is a growing need for automated, accurate, and scalable disease classification techniques that can support timely intervention. It introduces ASBOT (Adaptive Swarm-Based Optimization Technique), a hybrid machine learning algorithm that integrates Support Vector Machine (SVM) and Particle Swarm Optimization (PSO) for classifying guava leaf diseases. SVM is a powerful classifier but highly sensitive to its hyperparameters, especially the regularization constant C and the kernel parameter gamma (γ). ASBOT employs PSO to automatically optimize these parameters, thereby eliminating manual tuning and improving the model’s performance. By learning from color and texture features extracted from preprocessed leaf images, ASBOT demonstrates high accuracy and efficiency, offering a robust solution for automated plant disease diagnosis in agricultural applications.