Health Insurance Cost Predictor🫰

Project Overview

Situation: The project aims to help insurance providers estimate medical insurance payouts by predicting health insurance costs for individuals.
The goal is was to develop a machine learning model that accurately forecasts "claim" amounts to optimize premium pricing and financial risk assessment.

My steps:

• 1. Exploratory Data Analysis (EDA): Analyzed feature distributions and correlations, discovered that smoking status and BMI were the strongest predictors of high claim amounts
• 2. Feature Engineering: Created categorical bands for age and BMI (e.g., underweight, obese) to better capture non-linear relationships
• 3. Data Preprocessing: Handled missing values, applied Label Encoding for categorical variables, and used StandardScaler for feature scaling
• 4. Model Training & Optimization: Evaluated five regression models (Linear, Polynomial, Random Forest, SVR, and XGBoost) using Grid Search for hyperparameter tuning.
• 5. Deployment: Developed an interactive Streamlit web application that allows users to input their health data and receive real-time insurance cost estimates

Results:

• Best Model: XGBoost outperformed other models, achieving around 78% precision score.
• Outcome: Built a scalable end-to-end pipeline that automates the transition from raw data to a user-friendly prediction tool.