An Adaptive Multi-Stage Ensemble Learning Approach for Accurate Diagnosis of Tomato Leaf Diseases

Leaf diseases of tomatoes highly influence the quality and productivity of tomato plants; hence, it is important to diagnose the diseases correctly and in time to control them efficiently. Due to variations in ambient conditions, leaf texture, and disease symptoms, conventional approaches like deep learning and machine learning often fail generalization. Adaptive Multi-Stage Ensemble Learning (AM-SEL) for Tomato Leaf Disease Diagnosis is a system that integrates ML and DL models to address these challenges. It attains strong and accurate disease detection. There are three phases to the AM-SEL approach: (1) Feature Extraction &Preprocessing, where raw leaf images are augmented and hybrid features are extracted from CNNs and handcrafted feature descriptors; (2) Multi-Model Ensemble Classification, which enhances prediction accuracy through the integration of ML classifiers (Random Forest, Support Vector Machine) and DL models (ResNet, EfficientNet) and optimizes ensemble learning through a meta-learning strategy; and (3) Adaptive Decision Optimization, which minimizes misclassifications through maximum accuracy. Several performance metrics, including accuracy, precision, recall, and F1-score, are employed to evaluate this framework's applicability to large-scale tomato leaf disease datasets. The simulation results indicate that AM-SEL outperforms conventional single-model approaches regarding classification accuracy and robustness against image modifications. Due to its capability for real-time disease monitoring via edge computing and mobile-based diagnostic applications, the proposed method has practical applications in precision agriculture. Future studies will improve agricultural decision-making by adding Internet of Things (IoT) sensors and expanding the framework for detecting diseases in multiple crops simultaneously.