A Comparative Study of Machine Learning Algorithms for Plant Leaf Disease Classification Using Hybrid Feature

Plant leaf diseases are among the major factors contributing to reduced agricultural productivity and crop quality. Manual disease identification is time-consuming and requires considerable experience and accuracy, making it less efficient when applied on a large scale. Advances in image processing and machine learning offer opportunities to develop automated plant disease classification systems based on leaf images. This study aims to compare the performance of Support Vector Machine (SVM), Random Forest, K-Nearest Neighbor (KNN), and Naive Bayes algorithms in classifying plant leaf diseases using color, texture, and shape features. The dataset used in this study is the Plant Leaf Dataset, which consists of 3,695 leaf images across four classes: Tomato Healthy, Tomato Leaf Mold, Potato Healthy, and Potato Late Blight. Color features were extracted using Color Moments in the RGB and HSV color spaces, while texture features were extracted using the Gray Level Co-occurrence Matrix (GLCM) and Local Binary Pattern (LBP). Shape features were extracted using contour analysis and Hu Moments. To address class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) was applied. The experimental results show that SVM achieved the highest accuracy of 97%, followed by Random Forest and KNN, both at 95%, while Naive Bayes achieved an accuracy of 57.10%. These findings indicate that SVM is the most effective algorithm for plant leaf disease classification on the dataset used in this study.