A new secondary decomposition ensemble learning approach for carbon price forecasting

Abstract

The forecasting of carbon price plays a significant role in gaining insight into the dynamics of carbon market around the world and assigning quota about carbon emissions. Many studies have shown that decomposing the original data into several components with similar attributes is a widely accepted method addressing highly complex data. The resulting issue is that the high complexity of some components obtained is still tricky. This paper develops a new secondary decomposition strategy, which employs the complementary ensemble empirical mode decomposition (CEEMD) and the variational mode decomposition (VMD) to decompose the original series and the acquired intrinsic mode functions (IMFs) with maximum sample entropy value, respectively. All components are forecasted, including these generated by the first and secondary decomposition. The final results are obtained by synthesizing the predictions of all components. The experimental study states clearly that the established approach is superior to all benchmark models in terms of multistep horizons forecasting, and can provide the reliable and convincing results.