Interpretable machine learning for health insurance cost prediction based on health risk factors

Abstract

The utilization of machine learning in the healthcare domain holds immense potential to transform how individuals and organizations oversee the well-being of their workforce.Nevertheless, conventional machine learning models can present difficulties in interpretation and comprehension, posing challenges in explaining the model’s decisions and effectively integrating the acquired insights into policy and practice. For the purpose of this research proposal, we shall apply interpretable machine learning techniques on a dataset of employee health and wellness data.Our aim is to develop a model that can accurately predict health status, while simultaneously offering transparent and comprehensible explanations regarding the underlying factors influencing the predictions. To attain this objective, we will utilize an array of interpretable machine learning techniques and partial dependence plots. Additionally, we will employ methods like LIME and Shapley values to deconstruct the model’s predictions and assign them to specific input features. With this endeavor, our intention is to offer a robust tool for enhancing employee health while gaining a comprehensive comprehension of the elements that influence employee health outcomes. The outcomes of this project will be used to guide decisions on policies and practices concerning employee wellbeing, aiming to enhance the overall health and productivity of the workforce