Hybrid Deep Learning and Boosting Approach for Palm Tree Leaf Disease Classification

Abstract

This paper introduces a hybrid model for palm tree leaf disease classification by integrating Convolutional Neural Networks (CNN) with XGBoost. The proposed method exploits extract the features with multilayered CNNs, and that is coupled with the robust classification accuracy of XGBoost, to effectively identify diseases in palm tree leaves. The CNN model processes grayscale images, extracting complex features through multiple convolutional layers. These features are then used by the XGBoost classifier to accurately predict the presence of disease. Our experiments demonstrate that the hybrid CNN-XGBoost model achieves an accuracy of 97%, outperforming the standalone CNN model, which achieves an accuracy of 84%. This method significantly improves the precision and recall metrics, with the CNN-XGBoost model yielding an F1-score of 0.97 for the "Normal" class and 0.97 for the "Spotted" class.