This study investigate the efficacy of Random Forest Regressor (RFR) models as volatility prediction model applied in delta-neutral hedging strategies compared to traditional Historical Volatility (HV) and Implied Volatility (IV) approaches. 76 RFR volatility prediction models are trained on diverse historical windows using 10 rigorously selected features from 20 designed candidates, and systematic hyperparameter optimization. These models are empirically tested in delta-neutral hedging context 34 European call options from 2022-2024. The results implicate that RFR models achieve superior performance compared to traditional benchmarks, with significant improvement in RMSE while maintaining competitive MAE performance. Furthermore, a non-negligible amount of reduction in transaction costs and high win rates against traditional model proved RFR volatility prediction models cost-effective hedging performance. These findings provide compelling evidence that RFR volatility prediction model can outperform traditional models with better volatility prediction and efficient rebalancing across different market conditions. The demonstrated adaptability and efficiency of the RFR framework position it as a robust and practical tool for enhancing delta-hedging strategies in quantitative finance.
Zhanwang Zheng (Tue,) studied this question.