TOMATO DISEASE CLASSIFICATION USING ENSEMBLE LEARNING APPROACH

Abstract

Ensemble learning methods for supervised machine learning have become trendy due to their ability to accurately predict class labels with different learner methods. It offers efficient models for good predictive capability, which tend to be large and slight imminent into the patterns or structure in the data. In this research work, the ensemble learning technique is used as a classification task to accurately predict class labels. There are six different types of tomato diseases are predicted such as Anthracnose, Bacterial canker, Bacterial spot, Bacterial speck, Early blight and Late blight. Data are collected from the local market and the database is created with 600 images and 100 images for each disease. Various techniques like contrast enhancement in preprocessing, k-means is used for segmentation of tomato disease, color, statistical color features, color co-occurrence matrix and shape features are extracted and finally ensemble learning is used to accurately predicting tomato diseases. The experimental result shows that the proposed framework outperforms well in predicting tomato diseases using ensemble learning methods. An ensemble method for multi-class classification task is compared with various boosting and bagging ensemble methods.