Medical Insurance Cost Prediction Using ML

Medical insurance companies assess a variety of factors before setting premium costs — such as age, BMI, smoking habits, pre-existing conditions, and family history. Traditionally, actuaries estimate these costs manually using statistical models. Machine learning, however, can predict insurance charges much more accurately by learning from historical patient data, leading to fairer premium pricing and better risk assessment models for insurance companies.

Using structured datasets containing patient demographics, medical histories, and previous insurance charges, regression models like Linear Regression, Decision Trees, Random Forests, and Gradient Boosting can predict future charges. Feature engineering techniques like encoding categorical variables (region, smoker status) and scaling numerical features (age, BMI) improve prediction accuracy. This system can help insurers automate premium calculation and make data-driven pricing decisions.

Start by gathering a dataset containing patient features like age, gender, BMI, region, smoking status, number of children, and existing medical conditions. Preprocessing involves encoding categorical features and scaling numerical ones. Regression models are trained to predict continuous output — the insurance charges. Hyperparameter tuning ensures model optimization, and model evaluation uses RMSE (Root Mean Squared Error), MAE (Mean Absolute Error), and R² scores to measure prediction quality.

• Load insurance datasets with patient demographics and past charges from Kaggle or real-world sources.
• Preprocess data by encoding categorical features (region, smoker status) and normalizing numerical features (age, BMI).
• Train regression models like Linear Regression, Random Forest Regressor, or Gradient Boosting Regressor to predict insurance costs.
• Evaluate model performance using RMSE, MAE, and R² metrics, ensuring models generalize well on unseen data.
• Build a simple web app where users input patient data to get real-time insurance cost predictions instantly.

• Medical Cost Personal Dataset – Kaggle
• Building Regression Models with scikit-learn
• Deploy ML Apps with Streamlit