Performance Evaluation of Machine Learning Models for Crop Yield Prediction

Abstract

Agriculture, a fundamental pillar of worldwide sustenance, greatly benefits from precise yield projections, which provide effective allocation of resources and well-informed decision-making. This work focuses on the crucial task of predicting agricultural yields by conducting a thorough comparative examination of several machine-learning models. The examined models include Linear Regression, Random Forest, Extreme Gradient Boost (XGBoost), K-Nearest Neighbors (KNN), Decision Tree, and Bagging Regressor. The results demonstrate subtle variations in performance, with Linear Regression highlighting constraints in its ability to make accurate predictions. Ensemble approaches, namely: Random Forest and XGBoost, demonstrate remarkable accuracy, achieving almost 97% and R² ratings. This highlights their ability to effectively capture complex agricultural patterns. The findings of this research provide valuable suggestions for professionals in agriculture and machine learning, making it easier to choose reliable models for predicting crop yields. It is also recommended to optimize Random Forest and XGBoost for accurate production predictions in practical agricultural scenarios. Future studies may focus on using sophisticated optimization approaches and incorporating specialized domain knowledge to enhance the precision of agricultural production prediction.