Volume 1 - Issue 4, November - December 2025

Bacterial leaf stripe disease is an emerging threat to coconut (Cocos nucifera) cultivation, causing foliar damage that reduces photosynthetic efficiency and nut yield. Early and precise detection remains a challenge under field conditions due to overlapping symptoms with other leaf disorders. This study presents a machine learning–based framework to automatically detect bacterial leaf stripe from digital images of coconut leaves. Both classical and deep learning methods were explored, including Random Forest (RF), Support Vector Machine (SVM), and Convolutional Neural Networks (CNN). Image preprocessing, feature extraction, and model optimization were carried out using real-field image datasets. Results show that CNN models achieved a classification accuracy of 96.8%, outperforming traditional models. The integration of environmental variables such as temperature and humidity further improved prediction accuracy. The models explain ability was validated using Grad-CAM heatmaps, which accurately highlighted infected regions. The study demonstrates the feasibility of ML-based disease prediction and provides a foundation for developing portable diagnostic tools for sustainable coconut farming.