AI-powered stock closing price prediction using LightGBM on real Optiver Kaggle dataset.
🔗 Live Demo: optiver-trading.vercel.app
- 🧠 LightGBM model trained on 500K real NASDAQ order book data
- 📊 12 engineered features — WAP, Spread, Imbalance, Price Pressure
- ⚡ FastAPI backend with real-time prediction
- 📈 Feature Importance chart — interactive React dashboard
- 🎯 Validation MAE: 5.1552
| Layer | Technology |
|---|---|
| ML Model | LightGBM |
| Backend | FastAPI |
| Frontend | React + TypeScript |
| Charts | Recharts |
| Data | Optiver Kaggle Competition |
| Deployment | Vercel + Render |
# Backend
cd backend
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt
uvicorn main:app --reload
# Frontend
cd frontend
npm install
npm start- Open optiver-trading.vercel.app
- Enter order book features (bid_price, ask_price, imbalance, etc.)
- Click Predict
- View predicted closing price + Feature Importance chart
Snehal Thombare — GitHub
