An Explainable AI-Driven Web Platform for Crop Yield Prediction and Crop Recommendation using Soil and Environmental Parameters

Impact of climate change on agricultural productivity in India has been increasingly severe, compromising crop yield stability and food security. Prediction of crop yield before the start of harvest season can play an important role for farmers to make better-informed decisions on crop storage, distribution and market strategies. To this end, in this research effort, we propose an integrated, hybrid and explainable Machine Learning (ML)-based framework for crop yield prediction and crop recommendation. Our proposed model goes beyond the traditional models by introducing a hybrid ensemble model of Random Forest (RF) and XGBoost (XGB) along with temporal trend analysis for climatic variables in order to capture trend variability for each season. The input parameters considered include temperature, humidity, rainfall, soil nutrients (Nitrogen, Phosphorus and Potassium) and historical yield trends that are used to enhance the predictive performance. The proposed model incorporates Explainable AI (XAI) for transparent explanations of the predictions based on SHAP (SHapley Additive exPlanations). Our web-based GUI has been designed to be accessible to all farmers with minimum technical knowledge. The proposed system predicts crop yield and recommends the best crop based on environmental and soil conditions using classification models (RF, Support Vector Machines, k-Nearest Neighbors & XGB classifiers). Experimental results confirm that the proposed hybrid model performs significantly better than the other standalone models, by improving predictive accuracy, robustness and generalizability. This research effort contributes to precision agriculture by providing an integrated, explainable predictive analytics system to assist farmers in their decision making.