Machine Learning Classification Models Comparison for Crop Damage Identification

Abstract

Plants provide over eighty percent of the food consumed by humans and livestock. Its safety is threatened by diseases and pests. The main objective of the paper is to identify health status of crop state using machine learning classification models. We manage to compare eight approaches: Logistic Regression (LG), K-nearest neighbors (KNN), Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB), Gaussian Naive Bayes (GNB), Support Vector Machine (SVM) and Multilayer perceptron (MLP), on Indian agriculture dataset. We also used the Principal component analysis (PCA) and attempted to re-apply the same models. This paper includes an overview of these previously mentioned models, comparison between them according to data used, and a conclusion about which model leads to the best performance. Tests conducted show that the Extreme Gradient Boosting (XGB) is the best model for these data and problem for all cases then Decision Tree (DT), Logistic Regression (LR), and K-nearest neighbors (KNN). The last three models being very close in results.