Meteorological Factor-Based Tomato Early Blight Prediction Using Hyperparameter Tuning of Intelligent Classifiers

Abstract

Early blight is a severe disease which affects several plant species, including tomato plants. Weather parameters such as temperature, leaf wetness, soil moisture, and relative humidity play a vital role in the growth of diseases in plants. The current study analyses the effect of weather parameters on the development of early blight disease in tomato plants by utilizing traditional machine learning techniques. A real-time dataset TomEBD, comprising five weather parameters, has been employed. Three resampling techniques—Synthetic Minority Oversampling Technique(SMOTE), K-Means SMOTE(KM-SMOTE) and Support Vector Machine SMOTE(SVM-SMOTE)—have been used to balance the dataset. Five different machine learning classifiers—k-Nearest Neighbor(kNN), Support Vector Machine(SVM), Random Forest(RF), Artificial Neural Network(ANN), and Kernel Extreme Learning Machine(KELM)—have been used to classify a plant as healthy or diseased based on meteorological factors. The five classifiers are used on the imbalanced and three balanced datasets, resulting in 20 models. Hyperparameter tuning of all five classifiers has been done for optimization. The results indicate that out of the 20 models evaluated, the proposed model KELM-KM - KELM classifier on KM-SMOTE balanced data outperforms all others with a mean accuracy of 85.82%. A comparison with the existing studies shows that KELM-KM outperforms the state of the art without involving any complex feature extraction techniques. Therefore, it can be used to alarm the farmers for fungicide spray on diseased plants in conducive environments.