Development of an Ensemble Learning Approach for Soybean Yield Prediction using Satellite and Meteorological Data

Accurate crop yield estimation is important for many agricultural issues, including agricultural management, national food policies, and international crop trade. For this purpose, various methods are used to predict product performance, and the use of satellite images increases every day. Satellite remote sensing techniques that cover large areas continuously can help in more accurate assessment of crop yields. This research develops an optimal model for predicting soybean yield in the Midwest region of the United States. The ensemble learning hybrid model was tested using satellite images and meteorological data during the dominant growth period. In particular, the Golden Eagle Optimization (GEO) algorithm was used to adjust the hyper-parameters of the XGBoost model to provide the best possible configuration to improve accuracy. The results showed that the GEO-XGBoost model had good results for soybean crop (R equal to 0.9377 and RMSE equal to 0.2394 tons/ha). These results show that the optimized GEO-XGBoost model can provide accurate predictions for soybean yield under different weather conditions and can also be extended to predict other crops in the future.