Predictive Modelling of Insurance Premium for High-Risk Health Categories Using Wearable-Derived Health Scores and Machine Learning Approaches

Abstract

Supriya Sapa and Sharvari Tamane

The integration of wearable device data which helps in continuous health monitoring, with health insurance decision-making offers significant opportunities for personalized risk assessment and premium management. In this study, health scores were computed from wearable-derived features such as step count, calories burned, heart rate, and blood oxygen saturation using Machine Learning models like Support Vector Machine, Random Forest and Gradient Boosting. And then subsequently catego- rized into three levels: good, average, and poor. This study aims to develop and evaluate machine learning models for predicting premium updates specifically for policyholders classified under the poor health category, utilizing calculated health scores and existing insurance policy data. Machine Learning model, including Multiple Linear Regression (MLR), is applied to predict pre- mium updates based on health score and policy variables. The dataset comprised health metrics, demographic characteristics, and historical premium data for poor health category individuals. Model performance was evaluated using standard regression metrics including Mean Squared Error (MSE), R-squared values, RMSE and prediction accuracy. The results demonstrate that ML models can effectively capture the relationship between poor health conditions and insurance premium changes, enabling insurers to design more dynamic, fair, and personalized pricing strategies. This research highlights the potential of combining digital health tracking with predictive modelling to enhance actuarial practices and improve risk management in the insurance sector.

PDF