🏥 Insurance Claims ML Pipeline

A complete end-to-end machine learning project that predicts healthcare claims outcomes using synthetic data and XGBoost. This project includes:

---

📊 Exploratory Data Analysis (EDA) & Feature Engineering

• Data Generation: Synthetic patient-level claims data
• Missing Data: Handling missing values with imputation strategies
• Visualization: Distributions, correlations, and transformations (e.g., log transformation for skewed cost)

---

🧠 Model Training & Tuning

• Training Pipelines for Three Tasks:
  • Claim Cost Prediction (Regression)
  • Fraud Detection (Classification)
  • 30-Day Readmission Prediction (Classification)
• Technologies:
  • Models built using XGBoost and scikit-learn
  • Preprocessing pipelines combining imputation, one-hot encoding, and scaling
  • Hyperparameter tuning (GridSearchCV optional)

---

⚙️ Backend API

• Framework: FastAPI backend serving real-time predictions and retraining endpoints
• Documentation: Automatic interactive API docs via Swagger UI

---

🎨 Frontend Interface

• Platform: Multipage Streamlit app with dedicated pages for:
  • Home
  • EDA & Insights
  • Claim Cost Prediction
  • Fraud Detection
  • 30-Day Readmission Prediction
• Features:
  • User-friendly forms for single and batch predictions
  • Visualizations and clear insights

---

☁️ Deployment

• Containerization: Dockerfile provided for containerizing the FastAPI backend
• Hosting: Deployed on Render/Railway (backend) and Streamlit Cloud (frontend)

---

🚀 Installation

1. Clone the Repository:

   git clone https://github.com/yourusername/healthcare-claims-ml-pipeline.git
   cd healthcare-claims-ml-pipeline

2. Install Dependencies:

   pip install -r requirements.txt

3. Generate Synthetic Data:

   python data/generate_synthetic_data.py

   This command creates data/synthetic_claims.csv.

4. Train the Models:

   python notebooks/train_models.py

   This script will:

   • Load and preprocess the synthetic data.
   • Split data into training and testing sets.
   • Train pipelines for claim cost, fraud, and readmission.
   • Save the trained pipelines and preprocessor into the models/ folder.

5. Run the FastAPI Backend Locally:

   uvicorn backend.main:app --reload

   Then, open http://localhost:8000/docs to view the interactive API documentation.

6. Run the Streamlit Frontend Locally:

   streamlit run frontend/Home.py

   Use the sidebar to navigate between pages.

---

🌟 Features

• Real-Time Predictions:
  Get predictions for claim cost, fraud, and 30-day readmission based on user input.

• Comprehensive EDA:
  Detailed visualizations including distributions (with log transformations if needed), correlation heatmaps, and boxplots.

• End-to-End Model Pipelines:
  Preprocessing (imputation, encoding, scaling) combined with XGBoost models.

• API Integration:
  FastAPI backend with endpoints for predictions and retraining, including file uploads for batch operations.

• User-Friendly UI:
  Intuitive, multipage Streamlit interface with forms and downloadable results.

• Deployment Ready:
  Dockerfile for backend containerization, with deployment on Render/Railway and Streamlit Cloud.

---

🔗 Live Demo

👉 Try the live app here

---

📊 Model Performance

• Claim Cost Prediction:

  • Best Model: XGBoost Regressor
  • R² Score: e.g., 0.82
  • RMSE: e.g., ~$4,500

• Fraud Detection:

  • Best Model: XGBoost Classifier
  • Accuracy: e.g., 95%

• 30-Day Readmission Prediction:

  • Best Model: XGBoost Classifier
  • Accuracy: e.g., 88%
    (Parameters tuned using GridSearchCV and compared against multiple baselines.)

---

💡 Technologies Used

• Programming Language: Python
• Data Manipulation: Pandas, NumPy
• Visualization: Matplotlib, Seaborn
• Modeling: Scikit-learn, XGBoost
• Backend: FastAPI, Uvicorn
• Frontend: Streamlit
• Containerization: Docker
• Version Control: Git & GitHub

---

🤝 Contributions

This project is open for feedback, improvement, and collaboration. Feel free to fork, star, and open issues or pull requests!

---

📜 License

This project is licensed under the MIT License.

---