Semi-parametric regression model prediction method based on Empirical Mode Decomposition

摘要

Semi-parametric regression model prediction method based on empirical mode decomposition was studied in this paper.Firstly, basic idea of the empirical mode decomposition was introduced, and the improved algorithm was proposed to mitigate the end effect in the iterative shift process. Secondly, least squares method was employed to estimate the parameter /3 based on the trend component of empirical mode decomposition, and the non-parametric g(·) was estimated through building the AR models of the intrinsic mode functions. The vector matrix was computed by Yule-Walker method. Finally, time series prediction of two nonlinear systems was analyzed based on the semi-parametric regression model. The results show that the proposed model predictive method is fit for nonlinear and nonstationary time series estimate.