Enhancing Transparency in Smart Farming: Local Explanations for Crop Recommendations Using LIME
A Yaganteeswarudu, Saroj Kr. Biswas, V. Aruna, Deeksha Tripathi
Abstract
The integration of Explainable Artificial Intelligence (XAI) into smart farming enhances Crop Recommendation (CR) systems by improving transparency and understanding through local explanations. This paper discusses developing and implementing an XAI-enabled CR system using the Local Interpretable Model-agnostic Explanations (LIME) algorithm. The system builds user trust and supports informed agricultural decision-making by offering local explanations for individual records. Using LIME, the system elucidates the impact of various features on crop predictions, assigning positive or negative values to indicate their contributions. A positive LIME value marks a feature’s favorable impact, while a negative value indicates a detrimental effect. The paper highlights the system’s effectiveness with detailed crop explanations. For example, nitrogen, humidity, and rainfall positively influenced coffee, while for lentils, rainfall, nitrogen, and potassium were significant.