Health Insurance Cost Predictor Using Machine Learning

Health Insurance Cost Predictor

This is a health insurance premium prediction system made using machine learning. It uses a smart method with two models to give accurate results, up to 99.7% R² score.

It’s a good project for students, developers, or anyone who wants to learn how to use ML in real-life apps like insurance.

Project Details

Download New Real Time Projects :-Click here

About the Project

This project started as a standard regression task but evolved into a dual-model powerhouse after deep error analysis revealed age as a major factor in prediction variance.

Core Strategy

•  Young Model (Age ≤ 25) → Linear Regression
•  Older Model (Age > 25) → XGBoost Regressor
• Genetical Risk Factor: Game-changing feature that boosted overall prediction reliability.

Key Results

Young Users (Age ≤ 25)

• Model: Linear Regression + Genetical Risk
• Accuracy (R²): ~0.99 Train/Test

Older Users (Age > 25)

• Model: XGBoost + Genetical Risk
• Accuracy (R²): ~0.997 Train/Test/CV

Tech Stack

• Frontend: Streamlit
• ML Models: Linear Regression, XGBoost
• Tuning: GridSearchCV, RandomizedSearchCV
• Libraries: NumPy, pandas, scikit-learn, XGBoost
• Visualization: Seaborn, Matplotlib

Available Features

• ML model training & prediction scripts
• Advanced feature engineering (Genetical Risk)
• Pre-trained model notebooks for both age groups
• Ready-to-use Streamlit web app
• Residual & accuracy visualizations
• Test datasets included (.xlsx)
•  SourceOfWork.pdf: Full methodology & implementation breakdown