AbstractIn an attempt to combat global warming, many countries have been introducing carbon trading schemes, which, in turn, has led to increased research interest in carbon price forecasting as accurate predictions can provide valuable decision references for governments and investors. Therefore, this paper proposes an optimal hybrid forecasting framework to ensure accurate long‐term carbon price series predictions for which empirical mode decomposition (EMD) was used to decompose the original data and a novel optimal combination developed to predict each decomposed part, which involved the integration of autoregressive moving average‐type (ARMA‐type) models and least squares support vector machines (LSSVMs) optimized by 10‐fold cross validation and particle swarm optimization (PSO) algorithm. The effectiveness of the optimal EMD‐ARMAs‐LSSVMs framework was then verified using an empirical example from the EU emissions trading system (EU ETS). The long‐term forecast results and comparisons with other forecasting methods demonstrated the stability and validity of the proposed framework. The results indicated that ARMA‐type models are more suitable for low frequency and trend components, and LSSVMs were more suitable for high frequency components. It was also found that the proposed optimal framework was able to effectively reduce the error accumulation.
Read full abstract