XGBoost-Based Framework for Robust Crop Yield Prediction under Uncertainty Using Fuzzy Logic and Genetic Optimization

The precise forecasting of crop yield is vital in sustainable agricultural practices but it has not been achieved because of the vagueness that exists in the soil, climatic conditions and management practices. Therefore, the suggested work presents a powerful hybrid framework that is based on XGBoost, supplemented with a fuzzy logic and genetic optimization, to increase the accuracy of predictions in the case of uncertain conditions. The model combines Intuitionistic Fuzzy Sets (IFS) and Interval Type-2 Fuzzy Sets (IT2FS) to describe the vagueness and hesitation of the key features of the Area-Production-Yield (APY) dataset. Genetic Algorithm (GA) is used to optimize the selection of features, making them less redundant and enhancing their generalization. The system proposed offers a better performance than the baseline models having MAE of 0.0003, RMSE of 0.0010, R2 of 0.9994 and Fuzzy Consistency Index (FCI) of 0.9993. This framework is able to improve accuracy of prediction as well as facilitate interpretable decision making accompanied by uncertainty in precision agriculture. It can be scaled and modified to be implemented in the real-world in the various agro-climatic regions.