Health Insurance Premium Predictor
Machine learning–based premium estimation with real-world risk segmentation
Domain / Function
Healthcare Analytics & Predictive Modeling
Project Overview
Developed a machine learning system to accurately predict health insurance premiums based on customer demographics, lifestyle factors, and medical history. The solution addresses non-linear relationships that traditional pricing methods fail to capture.
The project follows an end-to-end ML pipeline including data cleaning, feature engineering, model training, error analysis, and deployment via a Streamlit web application for real-time predictions.
Premium Prediction Output 1
Premium Prediction Output 2
Premium Prediction workflow
Key Features
- High-accuracy premium prediction (R² ≈ 0.98)
- Advanced feature engineering using normalized risk scores
- Age-based model segmentation for improved accuracy
- Error analysis with residual and percentage deviation tracking
- Model retraining using additional genetic risk features
- Interactive Streamlit application for real-time use
Project Details
Multiple models including Linear Regression, Ridge Regression, and XGBoost were trained and compared. Initial error analysis revealed high deviation in younger age groups, leading to segmentation-based retraining.
After introducing a genetic risk feature, extreme prediction errors were reduced from 73% to just 2%, resulting in a reliable and explainable production-ready solution.