Skip to content

Latest commit

 

History

History
65 lines (50 loc) · 2.97 KB

README.md

File metadata and controls

65 lines (50 loc) · 2.97 KB
Raw

MedInsurance-Predictor

Project Overview

This project predicts healthcare insurance charges based on personal attributes, lifestyle factors, and geographic information. By analyzing features like age, gender, body mass index (BMI), family size, smoking habits, and region, the goal is to build machine learning models that accurately estimate insurance costs and provide insights into the key drivers of healthcare expenses.

Key Features:

  • Data Analysis and Visualization: Exploratory data analysis (EDA) to understand trends, relationships, and distributions in the data.
  • Feature Engineering: Transforming categorical variables, scaling features, and creating polynomial terms.
  • Machine Learning Models: Implementing linear regression, polynomial regression, and regularized models (Ridge, Lasso, ElasticNet) to predict healthcare charges.
  • Evaluation: Cross-validation, learning curves, and performance metrics like RMSE and R² to validate model accuracy and generalization.

Data Source

The dataset used in this project is publicly available on Kaggle: Healthcare Insurance Dataset.

Project Structure

  • data/: Contains the dataset files (train and test data).
  • notebooks/: Jupyter notebooks with the data analysis, model training, and evaluation steps.
  • src/: Python scripts for data preprocessing, feature engineering, model building, and evaluation.
  • results/: Model outputs, including learning curves, RMSE values, and predictions.
  • README.md: This file with an overview of the project.
  • requirements.txt: List of dependencies and Python packages used in the project.

How to Run the Project

  1. Clone the repository: 
    git clone https://github.com/Heinyxiao/MedInsurance-Predictor.git
cd healthcare-insurance-prediction
  2. Install the required dependencies: 
    pip install -r requirements.txt
  3. Open the Jupyter notebooks to run the analysis and model training: 
    jupyter notebook notebooks/MedInsurance_Predictor.ipynb
  4. Alternatively, you can run the Python scripts for training models: 
    python src/MedInsurance_Predictor.py
  5. Run the Interactive Web App: This project includes an interactive web app where users can input their personal data to predict healthcare insurance prices. 
    pip install streamlit
streamlit run app.py

Model Evaluation

The best-performing model in this project was Polynomial Regression (degree=2) with Lasso Regularization:

Test RMSE: 4,726.90

Test R²: 0.852

Future Improvements

Incorporating more advanced machine learning models (e.g., Random Forest, Gradient Boosting). Performing feature selection and experimenting with interaction terms. Further tuning regularization parameters for improved accuracy.

License

This project is licensed under the MIT License - see the LICENSE file for details.