Neema-ac/UTCPS-Nairobi

AI-powered urban traffic congestion prediction system for Nairobi, Kenya # UTCPS — Urban Traffic Congestion Prediction System ### Nairobi, Kenya 🇰🇪 An AI-powered platform that forecasts traffic congestion levels (Low, Medium, High) at least 30 minutes in advance across 12 key Nairobi road segments. ## Features - Real-time traffic data via TomTom Traffic API - LSTM/GRU model (TensorFlow) — Test RMSE: 0.1378 - GNN model (PyTorch) — Test RMSE: 0.2081 - Anomaly detection (accidents, road closures) - Interactive Streamlit dashboard with Nairobi map - REST API with 9 endpoints (FastAPI) - Weather impact analysis (OpenWeatherMap) - Docker containerization ## Tech Stack Python 3.11 | TensorFlow 2.15 | PyTorch 2.2 | Apache Kafka | Streamlit | FastAPI | Docker | SQLite ## Setup ### 1. Clone the repository git clone github.com cd UTCPS-Nairobi ### 2. Create virtual environment (Python 3.11 required) py -3.11 -m venv .venv .venv\Scripts\activate ### 3. Install dependencies pip install -r requirements.txt ### 4. Configure API keys Create a .env file in the project root: TOMTOM_API_KEY=your_key_here OPENWEATHER_API_KEY=your_key_here ### 5. Download dataset Download metr-la.h5 from: drive.google.com Save to: data/raw/metr-la.h5 ### 6. Preprocess and train python src/preprocessing/preprocess.py python src/models/lstm_model.py python src/models/gnn_model.py ### 7. Run the system # Terminal 1 - Dashboard streamlit run src/dashboard/app.py # Terminal 2 - API uvicorn src.api.api_server:app --reload --port 8000 ## Dashboard Login - Admin: admin / utcps2024 - Planner: planner / nairobi24 ## Project Structure UTCPS/ ├── src/ │ ├── ingestion/ # Nairobi roads, TomTom fetcher, Kafka │ ├── preprocessing/ # METR-LA processing, feature engineering │ ├── models/ # LSTM, GNN, anomaly detector, logger │ ├── api/ # FastAPI REST endpoints │ └── dashboard/ # Streamlit app, analytics page ├── config/ │ └── config.yaml …