club
Algorithmic Trading Predictor
An XGBoost classifier for predicting RELIANCE.NS price direction five trading days ahead.
Built an XGBoost classifier using stationary momentum and volatility features: log returns, PPO, normalized ATR, and Bollinger %B.
Implemented gap-5 purged time-series cross-validation to remove overlap with the five-day label horizon.
A 147-trade long/short backtest reported a 1.48 annualized Sharpe and 0.57 F1, assuming zero transaction costs. These are résumé-reported historical backtest results, not live trading performance or investment advice.
Context: Quant Club project.