Comparative analysis of Hybrid Machine learning Models for Crop Yield Prediction

Prediction of crop production ahead of harvesting season allows policymakers and farmers to plan for marketing and storage. Predictions must be accurate to optimize agricultural practices and enhance agricultural output. Traditional approaches for crop yield prediction rely on historical data, weather trends, and expert opinions but frequently fail to provide the accuracy and scalability required for sustainable agricultural growth. Our work explores the usability and applicability of different machine learning (ML) techniques for crop yield prediction. It addresses the issues and challenges of existing ML models, for their real-time implementation difficulties and low accuracy rates. Here we proposed three hybrid models that are built on top of popular machine learning models namely CatBoost Regressor, Extreme Gradient Boosting (XGBoost), Random Forest Regressor, and Ridge regressor to overcome these limitations.The performance of these hybrid models is evaluated for predicting crop yield and the comparative analysis of these models has also been carried out with the available traditional model. This initiative will help policymakers and farmers to predict crop yield before the harvesting season and enable them to make informed decisions.