Machine Learning Approaches to Predict Crop Yield Using Integrated Satellite and Climate Data

Abstract

India is the second-largest producer of wheat crop. Timely and appropriate prediction of wheat crop yield is essential for global and local food security. This research work has integrated multiple source data to predict crop yield across the Rajasthan state of India using lasso regression, support vector machine, random forest regression, and linear regression for crop yield prediction. This study used multiple vegetation indices (enhanced vegetation index, normalized vegetation index, soil adjusted vegetation index, chlorophyll vegetation index, and normalized difference water index). The results indicated that integrating multiple source data improves the model performance for all the machine learning models. Satellite data contributed additional information to the crop yield prediction than other data variables, and SAVI achieved better performance than other vegetation indices. We found that the support vector machine outperformed all the other approaches. The present study is a significant effort to integrate the multiple source data for the considerable area yield prediction.