Advanced Machine Learning Approaches for Accurate Sugarcane Yield Prediction with Ensemble Techniques

Sugarcane is a cornerstone of Bangladesh's agricultural economy and the global sugar industry, providing critical raw materials for sugar, biofuels, and various derivatives. Accurate yield prediction is essential for optimizing resources, ensuring economic stability, and fostering sustainable agriculture. This study harnesses advanced machine learning techniques to predict sugarcane yield, employing models like Decision Tree, Random Forest, Gradient Boosting, and K-nearest neighbors (KNN). It also explores ensemble learning approaches, including voting and stacking, to further refine predictive accuracy. Hyperparameter tuning significantly enhanced model performance, with Gradient Boosting and Random Forest achieving R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> scores of 99.38% and 99.24%, respectively. Notably, the Stacking Ensemble outperformed all individual models, delivering an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score of 99.24% and the lowest Mean Squared Error (MSE) of 7.20. Feature importance analysis using tree-based models uncovered key drivers of yield, providing actionable insights for precision agriculture. This work demonstrates the transformative potential of machine learning, particularly ensemble methods, in agricultural analytics. The study lays a foundation for scalable, data-driven decision-making in global farming practices by integrating predictive accuracy with interpretability.