Machine learning based classification of rice leaf diseases: A comparative study

Abstract

Rice is a staple food for most of the world’s population, making its sustainable production essential. However, undetected outbreaks of rice leaf diseases often result in reduced production due to crop failure. Accurate identification of rice leaf diseases is an important step in effective disease management. Machine learning (ML) algorithms can be an effective solution for early classification of rice leaf diseases based on available data. This study compares the performance of five ML algorithms, namely K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Random Forest, Decision Tree, and Naive Bayes, in classifying rice leaf diseases. The evaluation uses metrics such as accuracy, precision, sensitivity, F1 score, and training and prediction time. In addition, the analysis also includes ROC-AUC values and statistical tests such as Friedman test and Nemenyi post hoc test to determine the best algorithm significantly. The results showed that Random Forest had the best performance among the five algorithms with a ROC-AUC of 0.92 and an accuracy of 0.94, however, it is noteworthy that this particular algorithm requires a longer duration for both training and prediction compared to its counterparts. Ultimately, this study serves to provide invaluable insights and guidance for the selection of the most optimal machine learning algorithm, thereby facilitating more efficient and sustainable practices in the classification of rice leaf diseases, which is crucial for the advancement of agricultural productivity and food security.