A machine learning-based spray prediction model for tomato powdery mildew disease

摘要

Powdery mildew is the most commonly observed disease of tomato plants, which affects its quality and productivity. On-time treatment with an optimized amount of fungicides spray can improve the yield and quality of tomato fruit and reduce the effect of powdery mildew disease. Thus, an automated spray prediction model can be developed to deal with the tomato powdery mildew disease. In the current study, three spray prediction models have been proposed using different machine learning models viz. k -Nearest Neighbor ( k NN), Support Vector Machine (SVM) and Naïve Bayes (NB). This experiment was conducted using an imbalanced dataset named as tomato powdery mildew disease (TPMD) dataset. Synthetic minority over-sampling technique (SMOTE) has been used to balance the TPMD dataset. Results show that SVM based spray prediction model performed the best with 98.65% accuracy amongst all three models. An application version of spray prediction models can help the farmers for spraying fungicides in an efficient manner.