Machine Learning Driven Precision Agriculture: Enhancing Farm Management through Predictive Insights

Abstract

Machine learning (ML) is transforming all the fields  including agriculture, by providing data driven insights, which has enhanced and transformed the decision making process. Recent advancements in sensor technology, Wireless Communications, GPS and Data Analytics has led to widespread use by farmers which has led to increased resource utilization and efficiency and means to practice sustainable farming. In the proposed method various ML models are used to forecast two crucial metrics for precision agriculture Growing Degree Days (GDD) and Evapotranspiration (ET), which can be used for effectively managing daily agricultural activities. They are useful in predicting growth stages of crops, pest warnings, fertilizer usage, and irrigation times. Hourly data collected though sensors and other sources such as temperature, humidity, wind speed and soil moisture is used in managing real time growth of crops. The study assessed machine learning models like Random Forest, Support Vector Regressor (SVR), Voting Regressor, Stacking Regressor, and Decision Trees for precision agriculture metrics. Random Forest performed best but struggled with ET ranges. Decision Tree showed potential overfitting and underperformed, Voting Regressor and Stacking Regressor showed high performance. Despite hyper parameter optimization, the artificial neural network (ANN) exhibited poor performance, suggesting issues with either model selection or data adherence. The study developed a Decision Support System (DSS) that uses GDD and ET forecasts to provide real-time recommendations for pest and disease risk, fertilizer usage, crop maturation stages and watering regimens. The aim of this system is to equip farmers with the necessary tools to efficiently and effectively oversee their farms, hence enhancing agricultural productivity and sustainability.