An Improved EEMD-based Framework for CPI Forecasting

Although the Empirical Mode Decomposition (EMD)-based decomposition and ensemble framework for time series forecasting has been widely used, the end effect of EMD has not been addressed adequately. This study proposed to incorporate Mirror Method (MM), capable of dealing with the problem of end effect, into a hybrid modeling framework with Ensemble Empirical Mode Decomposition (EEMD) and Support Vector Machines (SVMs) for Consumer Price Index (CPI) Forecasting. The monthly Chinese CPI series from Jan. 2000 to Nov. 2011, with a total 143 observations, were employed to justify the performance of the proposed framework. The results suggested that it performed better than all the selected counterparts in terms of RMSE and SMAPE.