Crop yield forecasting through novel machine learning dynamic ensemble selection techniques

Crop yield prediction is a crucial aspect of agricultural planning and decision-making. This study utilizes a Kaggle dataset featuring State, Year, Season, Crop, Area, Production, etc., employing extensive data preprocessing and one-hot encoding of categorical features to enhance predictive performance. Initially, ML regression methods including Multiple Linear Regression, Random Forest Regression, XGBoost, and Decision Tree Regression (DTR) were applied. Later, Stacking Regressor and Voting Regressor were employed. Subsequently, dynamic ensemble techniques, specifically Voting Regressor, integrated these models, dynamically selecting the best-performing ones for prediction. Remarkably, achieving an R-squared value of 99%, the dynamic ensemble significantly improved prediction performance. The study showcases the efficacy of dynamic ensemble techniques in crop yield prediction, offering valuable insights for agricultural decision-making and resource allocation strategies. Experimental results underscore the effectiveness of the approach, highlighting its potential for practical application in agricultural management and policy formulation.