Explainable Artificial Intelligence for Crop Yield Prediction Using Agro-Climatic Data

Proper forecasting of crop yield is a requirement of proper agricultural planning and protection of food security. The current work presents a broad machine-learning-driven model of the prediction of crop yield, combining the range of regression models with strict evaluation criteria and elucidable artificial intelligence (XAI) methodologies. Agro-climatic, edaphic and nutrient related predictors were trained in decision tree, random forest, gradient-boosting, k nearest neighbors, support vector regression and deep neural networks. The evaluation of performance was conducted with the help of R2 and Root-Mean-Square-Error (RMSE), which showed that ensemble and deep -learning methods provide the best predictive values. Specifically, the highest accuracy was achieved with the Random Forest and Deep Neural Networks, which underscore their ability to estimate complex non-linear relationships among the dataset. Analysis of feature-importance identified temperature, humidity and soil quality as the most influential predictors of yield. Besides, lack of systematic bias and high generalisation was supported by error residual diagnostics. The explanations provided with SHAP also provided global and local answers to the decisions made by the model, thus increasing transparency and interpretability. Together, the proposed framework provides accurate, consistent and interpretable crop-yield predictions, providing priceless insights into data-driven agronomic decision-making and the promotion of sustainable farming systems.