A full-stack machine learning project that predicts and compares ride fares across multiple ride-hailing platforms including Ola, Uber, Rapido, and inDrive.
The system uses trained ML models to estimate ride prices dynamically based on ride parameters, traffic conditions, time of day, and weather data.
- Predicts fares for:
- Ola
- Uber
- Rapido
- inDrive
- Real-time feature engineering:
- Hour-based traffic logic
- Weekend detection
- Surge multiplier calculation
- Weather-based adjustment using latitude & longitude
- Platform-specific vehicle type mapping
- REST API built using FastAPI
- Rate limiting and caching implemented
- Large models hosted on HuggingFace Hub
- Frontend integrated and deployment-ready
- Built 4 independent regression pipelines using:
scikit-learnColumnTransformerOneHotEncoder(handle_unknown="ignore")
- Feature set includes:
- Distance
- Duration
- Hour
- Weekend flag
- Traffic level
- Weather
- Surge multiplier
- Passenger count (OLA)
Models are serialized using joblib and loaded dynamically.
- Built using FastAPI
- Lazy model loading
- IP-based rate limiting (token bucket)
- Prediction caching
- Weather API integration (Open-Meteo)
- HuggingFace model download integration
- Interactive ride comparison UI
- Displays predicted prices from all platforms
- Session-based result handling
- Deployment-ready configuration
- Python 3.11
- FastAPI
- scikit-learn
- NumPy
- Pandas
- Joblib
- HuggingFace Hub
- HTML / CSS / JavaScript
- Backend: Render
- Frontend: Vercel
- Model Storage: HuggingFace Hub
- User enters ride details.
- Backend generates dynamic features.
- Models predict platform-specific fares.
- Results are returned and displayed on the dashboard.
- Add deep learning models
- Add real-time traffic API integration
- Add historical ride analytics dashboard
- Optimize inference latency
Built as a production-oriented ML deployment project.