Predicting Crop Yield using Long Short-Term Memory, Integrated Gradients and Shapley Additive Explanations

Agriculture is the backbone of India, which plays a major role in the economy of a country. Agriculture is based on planting and harvesting crops over time. Crops should be sowed according to the environmental factors and appropriate lands. It is difficult for farmers to predict the yield of crop before harvesting because of varying climatic conditions. So to overcome this issue, the amount of crop yield is predicted with the factors like soil-type, weather, climate, crop-type etc., This paper introduces a novel approach to improve the accuracy and interpretability of crop yield prediction models by integrating Long Short-Term Memory (LSTM) networks, Integrated gradients (IG) feature selection, and SHapley Additive exPlanations (SHAP) analysis. The proposed methodology aims to predict the amount of specified crops that can be yield in an area. The lack of interpretability in many existing models hampers stakeholders' ability to understand and trust the predictions. The LSTM network captures temporal dynamics for improved accuracy, while IG and SHAP provide insights into feature contributions, addressing the interpretability issues in existing models. This integrated system not only advances prediction accuracy but also fosters understanding and trust among stakeholders, making it a comprehensive and interpretable tool for predicting crop yields.