Hybrid Deep Learning Framework for Automated Classification of Banana Leaf Diseases with Intelligent Anti-Agent Recommendation

The rapid spread of banana leaf diseases significantly affects crop yield and agricultural sustainability. This paper proposes a Hybrid Deep Learning Framework for Automated Classification of Banana Leaf Diseases with Intelligent Anti-Agent Recommendation, integrating convolutional neural networks (CNN) with attention-based feature enhancement for precise disease identification. The model leverages multi-level feature extraction combined with transfer learning to improve classification robustness under varying environmental conditions. Additionally, an intelligent recommendation module is incorporated to suggest appropriate anti-agents (fertilizers or pesticides) based on detected disease classes, enabling real-time decision support for farmers. The framework is trained and evaluated on a publicly available banana leaf dataset with extensive augmentation to enhance generalization. Experimental results demonstrate a classification accuracy of 99.86%, outperforming conventional deep learning models in terms of precision, recall, and F1-score. Furthermore, Grad-CAM visualization is used to ensure model interpretability by highlighting disease-affected regions. The proposed system offers a scalable, efficient, and farmer-friendly solution for early disease detection and management, contributing to smart agriculture practices and improved crop productivity.