Performance Evaluation of Machine Learning Techniques for Mustard Crop Yield Prediction from Soil Analysis

Abstract

Soil is an important parameter affecting crop yield prediction. Analysis of soil nutrients can aid farmers and soil analysts to get higher yield of the crops by making prior arrangements. In this paper, various machine learning techniques have been implemented in order to predict Mustard Crop yield in advance from soil analysis. Data for the experimental set-up has been collected from Department of Agriculture Department, Talab Tillo, Jammu; comprising soil samples of different districts of Jammu region for Mustard crop. For the current study, five supervised machine learning techniques namely K-Nearest Neighbor (KNN), Naïve Bayes, Multinomial Logistic Regression, Artificial Neural Network (ANN) and Random Forest have been applied on the collected data. To assess the performance of each technique under study; five parameters namely accuracy, recall, precision, specificity and f-score have been evaluated. Experimentation has been carried out to make known the most accurate technique for mustard crop yield prediction. From experimental results, it has been predicted that KNN and ANN (among the undertaken ML techniques for the study) found to be most accurate techniques for mustard crop yield prediction.