Forecasting volatility of China’s crude oil futures based on hybrid ML-HAR-RV models

Crude oil futures are central to global economic stability, with their volatility shaping financial markets worldwide. Forecasting volatility in China’s emerging crude oil futures market presents unique challenges, particularly during market stress events such as the COVID-19 pandemic and geopolitical disruptions. This study develops hybrid ML-HAR-RV models that integrate machine learning with econometric methods to enhance predictive accuracy and economic interpretability. Our analysis reveals pronounced jumps in volatility, with asymmetric responses to market shocks. Notably, the HAR-RV model incorporating signed jumps significantly improves predictive performance. Hybrid ML-HAR-RV models, especially those leveraging signed jumps, demonstrate superior forecasting capability. These findings refine the understanding of volatility dynamics in emerging futures markets and offer actionable insights for risk management and policy design. Beyond China, our framework provides a scalable approach for modeling commodity market volatility under external shocks, contributing to broader financial modeling and economic strategy.