Classification of Rice Leaf Diseases Using CNN-SVM Hybrid Model

Abstrak

Rice production in Indonesia faces significant challenges due to the declining area of agricultural land and the prevalence of diseases such as Bacterial Leaf Blight, Blast, and Brown Spot, which can reduce yields by up to 80% and threaten national food security. These diseases not only disrupt production stability but also cause substantial losses for farmers. Early identification is crucial to prevent such losses; however, farmers often struggle with limited knowledge, leading to misdiagnosis and improper management. To address this issue, this study proposes the development of a hybrid disease classification model for rice leaves based on Convolutional Neural Network (CNN) and Support Vector Machine (SVM). The model is designed following the CRISP-DM (Cross-Industry Standard Process for Data Mining) methodology, encompassing stages from business understanding to evaluation. Using a dataset containing 11,790 images of rice leaves from nine disease classes, the VGG-16 architecture is utilized for feature extraction, while SVM handles multi-class classification through the one-vs-rest approach. Evaluation results demonstrate a model accuracy of 95%, with high precision, recall, and F1-scores across most disease classes. These findings highlight the model's significant potential to aid farmers in the early detection of rice diseases.