Multi-Factor Volatility Prediction and Strategy Optimization for Quantitative Trading

Volatility is a key input for position control and risk management in quantitative trading. This study builds a multi-factor system that uses 14 market features—covering liquidity, reversal, and momentum—to forecast short-term volatility and adjust exposure. The model is trained with LightGBM in rolling windows, and its output is turned into a score for sizing positions. Tests on CSI 300 stocks show that this approach raises annual return by 3.7% and lowers the largest drawdown by 15% compared with a rule that relies only on recent volatility. The results show that a small set of market signals and a tree-based model can improve return and downside control. The method is simple to run and can be added to most trading setups. A main limit is that only daily data are used; adding intraday information may help during fast swings. Later work may test mixed-frequency inputs and regime-based sizing.