Health Insurance Cost Prediction API

A production-ready Machine Learning API built with FastAPI, containerized using Docker, and automated using CI/CD via GitHub Actions.

This service predicts health insurance expenses based on user inputs such as age, BMI, smoking status, region, and number of children.

---

Overview

This application exposes a REST API that serves predictions from a trained scikit-learn model.
The project follows production-oriented practices including:

• Modular architecture
• Automated testing with pytest
• Continuous Integration
• Continuous Deployment to Docker Hub
• Lazy model loading for runtime stability
• Dockerized deployment

---

Project Structure

Health_insurance/
│
├── src/
│   ├── api/
│   │   ├── main.py
│   │   └── schema.py
│   ├── data/
│   ├── features/
│   ├── models/
│   └── utils/
│
├── tests/
│   ├── test_api.py
│   └── test_train.py
│
├── Dockerfile
├── requirements.txt
├── requirements-dev.txt
├── pytest.ini
├── .github/workflows/
│   ├── ci.yml
│   └── cd.yml

---

API Endpoint

POST /predict

Request Body

{
  "age": 30,
  "gender": "male",
  "bmi": 25.5,
  "children": 1,
  "discount_eligibility": "no",
  "region": "southwest"
}

Legacy payloads with sex / smoker are also accepted for backward compatibility.

Response

{
  "predicted_expenses": 3456.78
}

Interactive API documentation:

http://127.0.0.1:8000/docs

Frontend UI:

http://127.0.0.1:8000/

---

Running Locally

1. Create Virtual Environment

python -m venv venv

Activate:

Windows

venv\Scripts\activate

Mac/Linux

source venv/bin/activate

2. Install Dependencies

pip install -r requirements.txt
pip install -r requirements-dev.txt

3. Run the API

uvicorn src.api.main:app --reload

---

Running with Docker

Build Image

docker build -t health-insurance-api .

Run Container

docker run -p 8000:8000 health-insurance-api

---

Running Tests

pytest

Tests are automatically executed in the CI pipeline on every push to the main branch.

---

CI/CD Pipeline

Continuous Integration

• Installs dependencies
• Runs unit tests
• Builds Docker image

Continuous Deployment

• Builds Docker image
• Pushes image to Docker Hub

Configured GitHub Secrets:

• DOCKERHUB_USERNAME
• DOCKERHUB_TOKEN

---

Model Training

To retrain the model:

python -m src.models.train

The model is loaded lazily at runtime to prevent import-time failures during CI execution.

If you uploaded a previously trained model and predictions fail, call:

curl http://127.0.0.1:8000/model-info

This returns the model's expected input columns so you can confirm schema compatibility without retraining.

---

Production Considerations

• Lazy model initialization
• Docker-based isolation
• Automated validation through tests
• Secure secret management
• Ready for cloud deployment

---

Docker Hub

Pull the production image:

docker pull <your-dockerhub-username>/health-insurance-api

---

Tech Stack

• Python 3.11
• FastAPI
• scikit-learn
• MLflow
• Docker
• GitHub Actions
• pytest

---

License

MIT License