Health Insurance Price Prediction Using Machine Learning Regression Models

Abstract

This paper presents a comprehensive Health Insurance Price Prediction system that employs six supervised machine learning regression algorithms to estimate individual insurance premiums with high accuracy. The system processes a dataset of 5,000 records characterised by six demographic and lifestyle features — age, sex, BMI, number of dependents, smoking status, and geographic region. Ridge Regression achieves the best generalisation performance with an R² score of 92.6%, a Mean Absolute Error (MAE) of $2,264.59, and a Root Mean Squared Error (RMSE) of $3,306.26 on a held-out test set of 1,500 records. The trained model is deployed as a Flask web application featuring secure user authentication, real-time prediction delivery (mean latency < 90 ms), personalised prediction history, and an interactive EDA dashboard. Statistical analysis confirms that smoking status is the dominant cost driver (Pearson r ≈ 0.79), followed by age (r=0.30) and BMI (r=0.20). Five-fold cross-validation (mean R²=92.5%, σ=0.8%) validates model robustness. Result analysis is supported by twelve graphs covering model performance, EDA patterns, risk distribution, and system benchmarks.