CLIMATE-AWARE MACHINE LEARNING FRAMEWORK FOR TOMATO DISEASE PREDICTION USING MULTI-ALGORITHMIC MODELS AND HYPERPARAMETER TUNING

Agricultural crop losses due to plant diseases pose a significant threat to global food security and farmer livelihoods, especially in climate-sensitive crops like tomatoes. In response, this study presents a climate-aware, multi-algorithmic machine learning framework for the early detection of tomato leaf diseases. Recognizing the critical role of climatic conditions—such as temperature, humidity, and rainfall—in disease emergence, the proposed approach integrates weather-based features to enhance prediction accuracy. The objective is to develop a robust and scalable solution for precision agriculture, capable of detecting multiple diseases across varying conditions. Four machine learning models were evaluated: XGBoost, Support Vector Machine (SVM), Random Forest, and a hybrid ensemble of SVM and Random Forest. Among these, XGBoost demonstrated superior classification performance, which was further improved using hyperparameter tuning via GridSearchCV. Tuning parameters like learning rate, number of estimators, and tree depth elevated XGBoost’s accuracy to 92% and yielded a weighted F1-score of 0.91. A detailed per-class analysis revealed excellent results for bacterial canker, fusarium wilt, and septoria leaf spot, though classes such as leaf spot and early blight remained more challenging due to sample limitations and visual similarity. Overall, the findings highlight the importance of integrating climatic data and tuning strategies in machine learning pipelines to enhance the reliability and generalizability of crop disease detection systems.