Diabetes Prediction Neural Network

A professional neural network system for diabetes prediction with a FastAPI backend and Streamlit frontend.

Demo

• https://diabetes-prediction-infnet.streamlit.app/

High Risk Prediction	Low Risk Prediction
Patient with high diabetes risk	Patient with low diabetes risk

Project Structure

diabetes-neural-network/
├── app/                          # FastAPI backend
│   ├── api/
│   │   └── v1/
│   │       └── endpoints/        # API endpoints
│   ├── core/                     # Configuration
│   ├── models/                   # Pydantic models
│   ├── services/                 # Business logic
│   ├── requirements.txt          # API dependencies
│   └── run.py                    # API runner
├── notebooks/                    # Jupyter notebooks
│   ├── baseline_model.ipynb      # Baseline model (89.57% recall)
│   ├── deeper_model.ipynb        # Best model (96.92% recall) ⭐
│   └── model_2025_09_21.ipynb    # Additional experiments
├── streamlit_app.py              # Streamlit frontend
├── streamlit_requirements.txt    # Frontend dependencies
└── start_services.py             # Start both services

Quick Start

Prerequisites

• Python 3.8 or higher
• pip (Python package installer)

Option 1: Start Both Services (Recommended)

1. Create and activate a virtual environment (Recommended)

# Create virtual environment
python -m venv venv

# Activate virtual environment
# On macOS/Linux:
source venv/bin/activate
# On Windows:
# venv\Scripts\activate

2. Install dependencies

# Install API dependencies
pip install -r app/requirements.txt

# Install Streamlit dependencies
pip install -r streamlit_requirements.txt

3. Start both services

python start_services.py

Option 2: Using the setup script

# Make setup script executable (macOS/Linux)
chmod +x setup.sh

# Run setup script (creates venv and installs dependencies)
./setup.sh

# Start services
python start_services.py

Option 3: Start Services Separately

Start API Server

# Make sure virtual environment is activated
source venv/bin/activate  # On macOS/Linux
# venv\Scripts\activate   # On Windows

cd app
pip install -r requirements.txt
python run.py

Start Streamlit App (in another terminal)

# Make sure virtual environment is activated
source venv/bin/activate  # On macOS/Linux
# venv\Scripts\activate   # On Windows

pip install -r streamlit_requirements.txt
streamlit run streamlit_app.py

Access Points

• Streamlit App: http://localhost:8501
• API Documentation: http://localhost:8000/docs
• API Health Check: http://localhost:8000/api/v1/health

Model Performance

Model	Architecture	Recall	File
Best Model	5-layer deep network	96.92%	deeper_model.ipynb ⭐
Baseline	2-layer simple	89.57%	baseline_model.ipynb
Experiments	Various architectures	~68%	model_2025_09_21.ipynb

API Endpoints

Health Check

• GET /api/v1/health - Basic health check
• GET /api/v1/health/detailed - Detailed health with model status

Predictions

• POST /api/v1/predict - Single patient prediction
• POST /api/v1/predict/batch - Batch predictions (max 100)
• GET /api/v1/model/info - Model information

Example API Usage

import requests

# Single prediction
response = requests.post("http://localhost:8000/api/v1/predict", json={
    "gender": 1,
    "age": 45.0,
    "hypertension": 0,
    "heart_disease": 0,
    "bmi": 25.5,
    "hba1c_level": 5.2,
    "blood_glucose_level": 120.0,
    "is_smoker": 0
})

print(response.json())
# Output: {"prediction": 0, "probability": 0.15, "confidence": "High"}

📊 Features

API Features

• ✅ RESTful API with FastAPI
• ✅ Input validation with Pydantic
• ✅ Automatic API documentation
• ✅ Health monitoring endpoints
• ✅ Batch processing support
• ✅ CORS enabled for frontend integration

Frontend Features

• ✅ Interactive Streamlit interface
• ✅ Real-time predictions
• ✅ Risk factor analysis
• ✅ Visual feedback and charts
• ✅ Professional UI/UX design

🛠️ Development

Best Practices

• Always use virtual environments to isolate project dependencies
• Keep dependencies up to date and pin versions in requirements.txt
• Test API endpoints using the interactive docs at http://localhost:8000/docs
• Follow PEP 8 coding standards for Python code
• Use type hints for better code documentation and IDE support

Adding New Features

1. API: Add endpoints in app/api/v1/endpoints/
2. Models: Update schemas in app/models/schemas.py
3. Services: Add business logic in app/services/
4. Frontend: Modify streamlit_app.py

Model Training

The best performing model is in notebooks/deeper_model.ipynb with:

• 5-layer architecture (10→64→32→16→1)
• Focal loss for imbalanced data
• 96.92% recall performance

Notes

• The API uses a placeholder model architecture for demonstration
• In production, load the actual trained model weights
• Configure CORS origins appropriately for production
• Add proper authentication and rate limiting for production use

Disclaimer

This tool is for educational and research purposes only. Always consult healthcare professionals for medical advice and diagnosis.