Improved Accuracy of Plant Leaf Classification using Random Forest Classifier over K-Nearest Neighbours

The aim of the study is to detect and classify healthy leaf and unhealthy leaf from the dataset provided by the Machine learning based novel Random forest algorithm. Materials and methods: A dataset of 2400 leaf images is created to train the random forest classifier which based on deep learning algorithms and is tested with 200 leaf images. Our proposed method includes steps of implementation such as dataset training creation and classification. The resulting statistics of affected and unaffected leaves simulated under random forest to categorize and separate the affected leaves and healthy leaf images with significance (p > 0.05). Result: On performing an independent samples t-test on the two similar groups, random forest achieved the accuracy, recall, precision and f1 score of 94.3%, 92%, 93% and 92% respectively compared to 87%, 84%, 81% and 83.5% by KNN algorithm. Conclusion: From this study, conclude the results of the detection in Random forest classifier appears to be higher than K-nearest neighbors.