Hybrid SVM-LR Classifier for Powdery Mildew Disease Prediction in Tomato Plant

Tomato plant suffers from various severe diseases; powdery mildew being one of them. Weather conditions play a significant role in the development of powdery mildew disease in tomato plant which in turn reduces the growth of tomato fruit. Hence, an accurate and timely detection of powdery mildew is necessary to extenuate the economic losses caused by the disease. This paper aims to develop a hybrid of Support Vector Machine (SVM) and Logistic Regression (LR) algorithm to predict powdery mildew disease in tomato plant. SVM is used to minimize the noise in data before the data is fed to LR classifier. Noise reduction is done using SVM classifier with the help of Adaptive Sampling based Noise Reduction (ANR) method. A real life Tomato Powdery Mildew Disease (TPMD) dataset has been used in this study to develop a prediction model using the proposed method. SVM and LR algorithms have also been used individually for developing the prediction models. Results indicate that the proposed classifier performs 3.06% better than SVM and 5.35% better than LR with an accuracy of 92.37%.