An intelligent transit analytics platform that visualizes historical TTC bus delays and predicts future incidents using machine learning.

• Real-time heatmap of TTC bus delays across Toronto (2014-2024)
• Time-lapse slider to traverse 10+ years of historical data
• Color-coded markers showing delay severity (green → yellow → red)

• Machine learning model trained on 160,000+ delay incidents
• Predict delays for different incident scenarios (mechanical, collision, emergency, etc.)
• Route-specific forecasting with confidence intervals

• Top delayed routes and stations
• Delay patterns by hour, day of week, and season
• Incident type breakdown and trends
• Real-time statistics

• Autocomplete station search
• Route comparison tool (compare up to 3 routes side-by-side)
• Location-based predictions

• React + Vite - Fast, modern UI framework
• Tailwind CSS - Utility-first styling
• Mapbox GL JS - Interactive map visualization
• Recharts - Data visualization library

• FastAPI - High-performance Python API
• LightGBM - Gradient boosting ML model
• Pandas - Data processing
• Scikit-learn - Feature engineering & calibration

• 160K geocoded delays (2014-2024)
• TTC GTFS data for stop locations
• 100+ engineered features (temporal, spatial, route-based)
• Isotonic regression calibration for accurate predictions

• Docker - Containerized deployment
• Render.com - Cloud hosting
• Nginx - Static file serving

• Python 3.11+
• Node.js 18+
• Mapbox API token (free tier works)

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

Backend runs at: http://localhost:8000

cd frontend
npm install
npm run dev

Frontend runs at: http://localhost:5173

Backend - No configuration needed!

Frontend - Create frontend/.env:

VITE_API_BASE_URL=http://localhost:8000
VITE_MAPBOX_TOKEN=your_mapbox_token_here

• Temporal: Hour, day of week, season, rush hour indicators
• Spatial: Distance from downtown, location frequency, spatial clustering
• Route: Route frequency, numeric route encoding
• Cyclical encoding for time-based features

• Algorithm: LightGBM (gradient boosting)
• Training: 5-fold time-series cross-validation
• Calibration: Isotonic regression for prediction accuracy
• Features: 85+ engineered features
• Data: 160,079 historical incidents (2014-2024)

• ✅ Removed data leakage (incident/spatial averages)
• ✅ Added prediction calibration
• ✅ More accurate delay estimates

GET /health

GET /api/historical?start=2024-01-01T00:00:00&end=2024-01-02T00:00:00

GET /api/predict?lat=43.7325&lon=-79.2631&route=36

GET /api/analytics/overview

GET /api/stations/search?query=kennedy

📖 Full API documentation: https://transight-backend.onrender.com/docs

docker build -f Dockerfile.backend -t transight-backend .
docker run -p 8000:8000 transight-backend

cd frontend
docker build \
  --build-arg VITE_API_BASE_URL=https://transight-backend.onrender.com \
  --build-arg VITE_MAPBOX_TOKEN=your_token \
  -t transight-frontend .
docker run -p 80:80 transight-frontend

Transight/
├── machineLearning/          # Backend API & ML model
│   ├── app.py                # FastAPI server
│   ├── predictor.py          # ML prediction engine
│   ├── feature_engineering.py # Model training
│   ├── ttc_delay_model.joblib # Trained model (5.5 MB)
│   └── requirements.txt
├── frontend/                 # React frontend
│   ├── src/
│   │   ├── components/       # React components
│   │   ├── api/api.js        # API client
│   │   └── App.jsx           # Main app
│   ├── Dockerfile
│   └── nginx.conf
├── dataAnalysis/
│   └── geocoded_delays.parquet # Historical data (2.9 MB)
├── Dockerfile.backend        # Backend container
├── render.yaml               # Render deployment config
└── README.md

• Minimalist UI - Clean, Apple/Stripe-inspired design
• Performance First - Optimized API calls, lazy loading, caching
• Accessibility - Keyboard navigation, semantic HTML
• Responsive - Works on desktop, tablet, and mobile

• TTC Delay Data: 2014-2024 historical delays from TTC open data
• GTFS Stops: Station locations from TTC GTFS feed
• Geocoding: ~70% match rate using stop name matching
• 160,079 incidents geocoded and processed

• Real-time delay alerts via WebSocket
• User accounts & saved routes
• Mobile app (React Native)
• Integration with TTC NextBus API
• Predictive routing suggestions
• Multi-city support

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

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

Built with ❤️ by a transit enthusiast who's tired of being late

• TTC for providing open data
• Mapbox for beautiful maps
• Render.com for free hosting
• Claude for code assistance and debugging

• 🌐 Live Demo: https://transight.onrender.com
• 📚 API Docs: https://transight-backend.onrender.com/docs
• 🐛 Issues: GitHub Issues

Made for Torontonians, by Torontonians 🍁