Comparative Analysis on Crop Yield Forecasting using Machine Learning Techniques

Abstract

Global overpopulation necessitates increased crop yields, yet available arable land is limited. The study compares and evaluates the performance of three machine learning algorithms—Random Forest (RF), Extra Trees (ET), and Artificial Neural Network (ANN)—in crop yield prediction. Using 28,242 samples with seven features from 101 countries, we evaluated these models based on Mean Absolute Error (MAE), R-squared (R^2), and Mean Squared Error (MSE). The ET regression model demonstrated superior performance, achieving an MAE of 5249.03, the lowest among the models tested. Despite having the highest R^2 value of 0.9873, the ANN exhibited higher MAE and MSE values, indicating less reliability. The RF model showed intermediate results. With a prediction accuracy of 97.5%, the ET model proved to be the most effective for crop yield prediction, achieving the highest accuracy reported to date. Future research should explore more advanced algorithms and larger datasets to validate these findings further.