Optimizing Crop Yield Prediction through Multiple Models: An Ensemble Stacking Approach

Abstract

Agriculture plays a pivotal role in enhancing India's economy, providing employment opportunities, supporting various industries, and contributing significantly to livelihoods and rural development. Accurate crop yield prediction is essential for effective crop management, productivity enhancement, and ensuring a balance between supply and demand. Leveraging machine-learning techniques, particularly stacking regressors, can offer improved predictive accuracy by capturing complex relationships between agricultural variables. This research paper conducts a comparative analysis of various machine learning techniques and introduces stacking ensemble learning for predicting crop yields in Indian agriculture. Each ML model, including Decision Tree, AdaBoost Regressor, and Linear Regressor, underwent individual training, testing, and prediction with hyperparameter tuning. Furthermore, the study implemented stacking ensemble learning using Linear Regressor, Decision Tree, and AdaBoost Regressor as base learners, with Linear Regressor serving as the meta-learner. The experimental results demonstrated that the stacking ensemble learning model outperformed all individual ML models, showcasing an impressive R2 value of 98.92%. These findings underscore the efficacy of stacking regressors in enhancing predictive accuracy for crop yield prediction, offering valuable insights for agricultural decision-making and resource allocation in the Indian agricultural sector.