•  
  •  
 

Abstract

An intelligent and explainable, as well as user-friendly, crop recommendation and yield prediction system is developed to facilitate agricultural decision-making under various tillage methods with the help of Explainable AI. The system utilizes Machine Learning algorithms and calculates yield changes as a relative term with respect to various inputs such as location, month of sowing, years of NT adoption, and soil type using real-time environmental data; accordingly, it recommends crops. The prediction framework, which is based on various regression algorithms, including Linear Regression, Decision Tree, Random Forest, and Gradient Boosting, was trained on the dataset of agronomic and climatic variables. Gradient Boosting gives better performance under data constraints, while Random Forest provides comparatively better performance for a larger dataset. For guaranteeing model interpretations and user faith, the system supports explainable AI methods like Shapley Additive explanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) for global and local interpretations of feature influence. The web-based interface facilitates smooth interaction and dynamic input handling based on weather and geolocation APIs. All inputs from the user along with the prediction results are stored in the backend database for monitoring and future optimization. The solution provides a systematic, interpretable, and functional approach to increase crop planning and sustainable land management by data-driven insight. The proposed approach helps farmers in achieving crop yields.

Download

Share

COinS