Comparative Evaluation of Machine Learning Models in Forecasting Crop Yields Amid Climate Change

This chapter delves into the comparative evaluation of machine learning models for predicting crop yields amid climate change, focusing on the performance of Multiple Linear Regression (MLR), Random Forest (RF), and XGBoost. The study highlights the importance of accurate yield prediction for global food security and explores how traditional statistical methods often fall short in capturing nonlinear climate-crop interactions. The research introduces advanced machine learning techniques, including XGBoost and Convolutional Neural Networks (CNNs), which excel in handling high-dimensional datasets and extracting spatial features from remote sensing data. The methodology involves a multi-step preprocessing pipeline, including anomaly-based feature engineering and dual normalization strategies, to enhance model performance and generalization. The results demonstrate that XGBoost outperforms other models, achieving the highest test accuracy and lowest Mean Squared Error (MSE), particularly when data normalization and augmentation are applied. The study also identifies key climate variables, such as the aridity index and CO2 emissions, as significant predictors of crop yield. Practical implications include the adoption of adaptation strategies like crop rotation and drought-resistant crops, which have shown to mitigate yield losses during climate stress. The chapter concludes by emphasizing the need for further validation with diverse global datasets and the integration of remote sensing data to provide deeper, actionable insights for resilient agricultural strategies.