Predicting Medical Insurance Costs: A Machine Learning Approach for Smarter Risk Assessment
Dontha Madhusudhana Rao, Lalith Kumar Raju Somalaraju, Venkata Sai Mahesh Bodagala, Venkata Bala Jyothi Swaroop Edapalapati · 2025
Rising healthcare costs make accurate prediction of insurance premiums essential for both insurers and policy-holders. In this study, we apply machine-learning regression models to forecast individual premiums using demographic and lifestyle factors. We used a dataset of 2,773 records and, during preprocessing, performed missing-value imputation to fill data gaps, removed outliers to improve model robustness, applied log transformations to reduce skew in continuous variables, and encoded categorical features into numeric form. After evaluating multiple algorithms, the XGBoost regressor demonstrated the highest performance with an R2of 0.9552 and a 10-fold cross-validation score of 0.9373, indicating both exceptional accuracy and robust generalization. We utilized SHAP (SHapley Additive exPlanations) to interpret the model outputs, which identified smoking status, age, and body mass index (BMI) as the key factors influencing insurance costs. These results provide critical insights for refining risk assessment, pricing strategies, and consumer decision-making. Future research may incorporate additional health metrics to further improve predictive accuracy.