Feature Importance in Health Insurance Data Using Variance-Based Sensitivity Analysis

Abstract

Variance-based sensitivity analysis serves as a crucial tool for assessing the variability of inputs on the output of complex mathematical models. In this study, we examine Sobol indices, a class of variance-based sensitivity analysis, to quantify the importance of each input variable on the overall variability of the model output. We apply Sobol indices within the framework of a regression model to identify the most importance features (also known as predictors) in predicting total medical expenses charged per year for the health insurance plan. Our findings reveal that “smoker” emerged as the most important feature impacting health insurance charges, followed by “age” and “bmi” as the second and third most important features, respectively. This application not only demonstrates the effectiveness of Sobol’s indices in regression models but also suggests possible model simplifications.