Enhancing Agricultural Forecasting with an Ensemble Learning Approach for Broccoli Yield Prediction

Abstract

Accurate yield prediction for broccoli remains a critical challenge due to the complex, non-linear interactions between climatic variables, soil properties, and agronomic practices. Traditional statistical models often fail to capture these interactions, leading to suboptimal decision-making and resource inefficiencies for farmers. To overcome these limitations, this study proposes an advanced machine learning-based approach called the Broccoli Yield Prediction Ensemble Method (BYPEM), designed to improve prediction accuracy and agricultural planning. BYPEM integrates both bagging and boosting ensemble learning techniques for robust broccoli yield forecasting. The study begins with a comprehensive data preprocessing phase, including handling missing values, outlier removal, categorical variable encoding, and normalization. Feature selection is performed using backward elimination to retain the most relevant predictors. The dataset is split into training and test sets through stratified sampling to ensure balanced representation. In the model development phase, BYPEM applies bagging methods such as Random Forest Regressor and Extra Trees Regressor to reduce variance, and boosting methods such as Gradient Boosting Regressor (GBR), XGBoost, LightGBM, and CatBoost to minimize bias by iteratively improving predictions. Hyperparameter tuning further optimizes model performance. The models are evaluated using multiple metrics including Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), R² score, and overall accuracy. Among all models, XGBoost achieves the highest predictive performance, confirming the effectiveness of the BYPEM framework in capturing complex yield dynamics. This research demonstrates how ensemble learning can support sustainable agriculture by enhancing precision in broccoli yield prediction and guiding data-driven farm management strategies.