Explainable Model for Agricultural Crop Yield Prediction in Indian Conditions with SHAP Analysis

Abstract

This paper presents an ensemble-based approach to agricultural crop yield forecasting, focusing on the Indian context. The study integrates different forecasting models to improve forecasting accuracy for agricultural complexity data analysis. SHAP (SHapley Additive exPlanations) is used to give a clear idea of the contribution of each factor in the prediction model to improve model interpretation. The dataset used for this study contains yields of 55 crops over 6 seasons in 30 countries in 23 years (1997)-2020 available). Besides demonstrating the method's effectiveness, it emphasizes the need for explicit modeling that can provide valuable insights for better agricultural practices and ultimately contribute to higher yields that will be sustainable in Indian agriculture.