Prediction modelling method for non-linear and non-stationary time series

A prediction modelling method was proposed for non-linear and non-stationary time series, based on empirical mode decomposition (EMD) and support vector regression (SVR). The time series was decomposed into several intrinsic mode components (IMCs) via EMD so as to make every component stationary. Then in view of the stationary time series, a prediction model was developed correspondingly for each and every IMC on SVR basis, and these prediction models were non-linearly combined together by use of SVR again to form the final prediction model for non-linear and non-stationary time series. Both simulative experiment and engineering application showed that the proposed method has higher precision in comparison with the conventional SVR-based modelling method, i.e., effective to non-linear and non-stationary time series prediction.