Early Prediction of Crop Yield in Bangladesh Using Ensemble Learning

Farmers in Bangladesh face financial losses due to difficulties in choosing optimal crops, influenced by factorslike pesticides, fertilizer, and rainfall. To address this, early crop yield prediction is crucial, particularly for rice, sugarcane, potato, and maize. This research employs machine learning algorithms-KNearest Neighbors, Random Forest, Gradient Boosting, AdaBoost, CatBoost, Decision Tree, and XGBoost-to evaluate their predictive performance. K-Nearest Neighbors, Random Forest, and Gradient Boosting emerged as the top performers. Advanced ensemble learning techniques, including stacking with various base and final estimators using these three models, are then explored to improve predictive accuracy. Rigorous evaluation metrics are applied, with maize yield prediction achieving $R^{2}$ 0.99 and MAE 0.04, rice $R^{2} 0.98$ and MAE 0.04, potato $R^{2}$ 0.99 and MAE 0.11, and sugarcane $R^{2} 0.99$ and MAE 0.36. This research advances agricultural prediction and demonstrates the effectiveness of ensemble learning in optimizing machine learning models for complex tasks. The findings provide stakeholders in agriculture with reliable tools for early crop yield estimation, essential for effective resource management and decision-making.