README (Updated with Streamlit & Algorithm Info)

Insurance Premium Prediction 🚀

This project is an end-to-end Machine Learning application that predicts insurance premiums based on user input (age, BMI, smoking status, region, etc.). It integrates a trained Random Forest Regressor model with a FastAPI backend, a Streamlit frontend, and is fully containerized using Docker for deployment on AWS EC2.

---

📂 Project Structure

insurance-premium-prediction/
│── main.py                 # Entry point for starting backend
│── app.py                  # FastAPI application (model serving API)
│── frontend.py              # Streamlit app (frontend UI)
│── model.pkl               # Pre-trained Random Forest Regressor model
│── insurance.csv           # Dataset used for training
│── patients.json           # Example input JSON
│── fastapi_ml_model.ipynb  # Notebook for experimentation & training
│── requirements.txt        # Python dependencies
│── .gitignore
│── Step_1_Creating_model_pkl.png
│── Step_2a_Creating_endpoint_API_Endpoint.png
│── Step_2b_Input_and_Output.png
│── Improved API.png

---

⚙️ Workflow Explanation

Step 1 – Model Training

• Dataset: insurance.csv
• Algorithm: Random Forest Regressor (trained in fastapi_ml_model.ipynb)
• Features: age, sex, BMI, children, smoker, region.
• Target: insurance premium (charges).
• Model saved as model.pkl for inference.

Step 2 – Backend (FastAPI)

• app.py loads model.pkl and provides /predict API endpoint.

• Workflow:

  1. Receive JSON input.
  2. Preprocess input.
  3. Call Random Forest Regressor model.
  4. Return premium prediction.

Step 3 – Frontend (Streamlit)

• frontend.py is a Streamlit app.
• It provides an interactive web interface where users can input values (age, BMI, smoking status, etc.).
• The Streamlit app sends data to the FastAPI backend and displays predictions in real-time.

Step 4 – Dockerization

• Dockerfile builds a containerized version of the app:

  • Installs dependencies from requirements.txt.
  • Runs FastAPI and Streamlit inside the container.

• Commands:

  docker build -t insurance-premium-prediction .
  docker run -p 8000:8000 insurance-premium-prediction

Step 5 – AWS Deployment

• Image uploaded to DockerHub.

• EC2 Ubuntu instance pulls image and runs container.

• Accessible at:

  http://<EC2-Public-IP>:8000/docs   (FastAPI)
  http://<EC2-Public-IP>:8501        (Streamlit)
  

---

🧩 How Code Works Together

• frontend.py (Streamlit) → User inputs data → sends request to FastAPI backend.
• app.py (FastAPI) → Receives request → uses Random Forest Regressor model (model.pkl) → returns prediction.
• Streamlit UI updates with predicted premium.

Flow: Streamlit → FastAPI → Random Forest Regressor model → Prediction → Streamlit Output

---

▶️ Run Locally

1. Install dependencies:

   pip install -r requirements.txt

2. Start FastAPI backend:

   uvicorn app:app --reload

   Available at http://127.0.0.1:8000/docs

3. Start Streamlit frontend:

   streamlit run frontend.py

   Available at http://localhost:8501

---

▶️ Run with Docker

1. Build Docker image:

   docker build -t insurance-premium-prediction .

2. Run container (FastAPI + Streamlit):

   docker run -p 8000:8000 -p 8501:8501 insurance-premium-prediction

3. Access:

   • API: http://localhost:8000/docs
   • UI: http://localhost:8501

---

▶️ Run on AWS EC2

1. Launch Ubuntu EC2 instance and install Docker:

   sudo apt update
   sudo apt install docker.io -y

2. Pull image:

   docker pull <your-dockerhub-username>/insurance-premium-prediction

3. Run container:

   docker run -d -p 8000:8000 -p 8501:8501 <your-dockerhub-username>/insurance-premium-prediction

4. Access:

   • http://<EC2-Public-IP>:8000/docs
   • http://<EC2-Public-IP>:8501

---

✅ Conclusion

This project demonstrates a complete ML pipeline:

• Random Forest Regressor model for premium prediction.
• FastAPI backend for model serving.
• Streamlit frontend for interactive UI.
• Docker for containerization.
• AWS EC2 for cloud deployment.

---