A novel decomposition-ensemble approach to crude oil price forecasting with evolution clustering and combined model

In order to deal with non-stationary and chaotic series, a hybrid forecasting approach is proposed in this study, which integrates ensemble empirical mode decomposition (EEMD) and optimal combined forecasting model (CFM). The proposed approach introduces a new intrinsic mode functions (IMFs) reconstruction method by using evolutionary clustering algorithm, and utilizes optimal combined model to forecast sub-series. Firstly, the EEMD technique is employed to sift the IMFs and a residue. Secondly, the comprehensive contribution index (CCI) of each IMF is calculated and IMFs are further reconstructed by evolutionary clustering algorithm according to CCI of each IMF. Then, a new sub-series called virtual intrinsic mode functions (VIMFs) is defined and obtained. Thirdly, the optimal combined forecasting model is developed to forecast the VIMFs and residues. In the end, the final forecasting results are obtained by summing the forecasts of VIMFs and residue. For illustration and comparison, the West Texas Intermediate (WTI) crude oil price data are shown as a numerical example. The research results show that the proposed approach outperforms benchmark models in terms of some forecasting assessment measures. Therefore, the proposed hybrid approach can be utilized as an effective model for the forecasting of crude oil price.