Forecasting realized volatility of Chinese crude oil futures with a new secondary decomposition ensemble learning approach

•We study realized volatility in the Chinese crude oil futures market.•A novel VMD-ICEEMDAN-LSTM prediction model is used to forecast realized volatility in crude oil futures market.•The learning approach model can provide better predictive performance and reliability than conventional econometric models.•Applying ICEEMDAN to the secondary decomposition of the VMD residual values can effectively improve the model's overall predictive ability.•Via MCS test, our VMD-ICEEMDAN-LSTM models outperform in forecasting realized volatility.