Explainable Hybrid Machine Learning Models for Soil Fertility Prediction with Generative AI-Based Crop Recommendation

Soil fertility evaluation is one of the key aspects for improving agricultural productivity and developing sustainable agriculture practices. Soil fertility evaluation is essential for farmers to take appropriate decisions regarding crop cultivation. In this research work, we propose an intelligent decision-support framework for farmers based on machine learning, explainable AI, and generative AI for soil fertility evaluation and crop cultivation. In the proposed framework, key soil parameters such as Nitrogen (N), Phosphorus (P), Potassium (K), pH level, moisture, temperature, organic carbon, and rainfall are considered as input features for predictive modeling. Three different machine learning algorithms, namely Random Forest, XGBoost, and Support Vector Machine, are implemented for soil fertility evaluation. Experimental evaluation of these algorithms is carried out for soil fertility evaluation, and it is observed that XGBoost algorithm produces 97% accuracy for soil fertility evaluation, whereas Random Forest and Support Vector Machine produce 92% and 72% accuracy, respectively. To improve the explainability of the proposed framework, explainable AI is implemented based on SHAP (SHapley Additive exPlanations), where the contribution of each soil parameter is analyzed for soil fertility evaluation. This would facilitate a better understanding of the impact of different soil characteristics on fertility classification. Additionally, a generative AI-based advisory module is proposed for incorporating context-based crop and farming recommendations based on the soil fertility and other conditions. This would enable the generation of customized recommendations for farmers regarding crops, fertilizers, and other soil management practices. Thus, with the integration of predictive ML-based approaches, explainable AI-based approaches, and generative AI-based approaches, the proposed framework would facilitate an intelligent decision-support framework for farmers with the aim of improving crop productivity and promoting sustainable farming practices.