Vol. 12, Issue 3 (2023)

Author(s):
Pawan Choudhary, Upendra Singh, US Tanwar and Vijay Singh Meena

Abstract:
Crop diseases are major issues as it reduces the production and quality considerably and creates a major threat to food security. But due to the lack of the necessary infrastructure and knowledge, immediate identification of diseases and rectification is not being done by most of the farmers which results in heavy crop damage and hence loss to them. This paper focuses on finding out the crop disease using data mining Classification techniques based on the physical characteristics of the crop. Classification algorithms like Decision Tree, Logistic Regression, Naive Bayes, Kernel SVM, kNN and Linear SVM are deployed on plant dataset. The performances of the algorithms are analyzed based on certain metrics like Execution Time, Accuracy Score, Cohen’s Kappa, Hamming Loss, Explained Variance Score, Mean Absolute Error, Mean Squared Error and Mean Squared Logarithmic Error. For analysis, Confusion matrix and Classification report are used. The Decision Tree is found to be the fastest algorithm and results in the best ATR. But the Linear SVM is found to be the best algorithm for all other metrics for the crop disease dataset. Various oversampling techniques like Synthetic Minority Oversampling Technique (SMOTE), SVM SMOTE and borderline SMOTE with sample and resample are applied to improve the performance of the Linear SVM. Linear SVM on the oversampled dataset using SVM SMOTE has a better performance for all metrics except Execution Time than that on the original dataset.