Evaluating Machine Learning Models for Wheat and Rice Disease Detection with Growth Optimization in Agriculture

One of the most important aspects of agriculture that helps to minimize output losses and maintain food security is the early detection of plant diseases. We performed data augmentation on two datasets of rice and wheat that we gathered from different sources to increase the size of the dataset for this study. Next, we used data pre-processing techniques such picture segmentation, feature extraction, color extraction, and shape extraction to extract the important features. Next, we trained the machine learning algorithms for XGBoost, Random Forest, K-Nearest Neighbor, Naive Bayes, and Ridge Classifier. Using the performance matrix which considers f1-score, accuracy, recall, and precision, we evaluated the algorithms' performance. With an accuracy of 95.83% for rice disease identification, Random Forest outperformed XGBoost and Random Forest with an accuracy of 91.66% for wheat disease detection, according to our experiments. According to the findings, Random Forest was the best at detecting all classes of rice illness, while KNN was the best at detecting the septoria class and Random Forest was the best at detecting the healthy and black rust classes of wheat disease. The information gathered can be used to create an automated system for classifying and detecting plant diseases, which may raise agricultural productivity and improve the availability of food.