Release Time:2019-03-12  Hits:

Indexed by: Conference Paper

Date of Publication: 2006-06-14

Included Journals: Scopus、CPCI-S、EI

Volume: 1-12

Page Number: 3698-+

Key Words: time series; NN; PCA

Abstract: This paper presents a new methodology for multivariate time series modeling and prediction. Different variables of the multivariate time series affect each other, so it is difficult to model multivariate system. A new technology is proposed consisting of Neural Networks (NN) and Principle Component Analysis (PCA). PCA is often used in the analysis of multivariate process data to identify important combinations of the original variables on which to focus for more detailed study. In this paper, it is applied to eliminate the redundant information in complicated and high dimensional input data stream so that the useful information can be processed in low-dimensional space. The proposed method is validated on multivariate time series prediction problems: the time series x(t) and z(t) of Rossler's equation; time series of annual sunspots and runoff of the Yellow River. Different simulations show the probability and validity of the proposed method.