Explanation of Prediction of Machine Learning Models Using Local Interpretable Model-Agnostics Explanation and SHapley Additive exPlanations With Case Study of Credit Card Defaulters

Artificial intelligence (AI) models, including machine learning and deep learning, play a crucial role in decision-making and predictive analysis across various fields. Such models often act as black boxes, as the basis of prediction is not provided. At times, understanding the basis of prediction is essential to assess the reliability of the model. To address this issue, the Explainable AI (XAI) technique is found to be useful. XAI helps in understanding how and why a model makes a certain prediction. This research paper focuses on feature-based techniques of XAI - SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-Agnostic Explanations) - which help in identifying the key features influencing the decisions of prediction models. These techniques are applied to two models, Random Forest and Decision Tree, developed in Python using a dataset from a case study titled Credit Card Defaulter, which identifies good or bad credit applicants. Two random scenarios are used to check the reliability of the prediction process. It has been observed that out of 12 features, the features code_gender, flag_own_car, flag_own_reality, and cnt_fam_members contributed the most to predicting good or bad credit applicants. The SHAP and LIME techniques used in explanation of prediction enhance the trust on the models used and identify the major contributing factors influencing the predictions.