Interpretation of machine learning models using XAI - A study on health insurance dataset

This study aims to dive into the complex Machine Learning algorithms using XAI and provide explanations for outcomes that we received from these algorithms. In this paper, we have proposed to predict the cost of health insurance. The proposed work is composed of two Machine Learning algorithms, namely Multiple Linear Regression and Random Forest, followed by an explanation of predicted results using XAI. Here, we first provide a simple explanation with the help of model-specific approaches based on Microsoft's InterpretML library. Then the predicted insurance premium cost is further explained by model-agnostic techniques, LIME, and SHAP. The significance of the study is to provide a better user experience and build trust between the user and the machine. These techniques can help to check the correctness of the prediction models, as the domain experts can analyze the features that affect the outcome the most and provide their expertise.