Title :
Markov gated experts for time series analysis: beyond regression
Author :
Shi, Shanming ; Weigend, Andreas S.
Author_Institution :
Dept. of Comput. Sci., Colorado Univ., Boulder, CO, USA
Abstract :
Most traditional time series models are based on local methods (in time), which means assuming that the time series can be fully and locally (in time) characterized with a finite embedding space. There are many situations in which simple regression can not help find the temporal structural in time series. In this research, a Markovian architecture, Markov gated experts, has been developed based on nonlinearly gated experts. This paper discusses the statistical framework and compares the performance of Markov gated experts to gated experts on both computer generated time series and real world data. Compared with the original method, Markov gated experts are more powerful in finding the underlying temporal structure, and are therefore a more powerful analytical and forecasting model for nonstationary and structurally changing time series
Keywords :
Markov processes; neural nets; time series; Markov gated experts; Markovian architecture; analytical model; finite embedding space; forecasting model; nonlinearly gated experts; nonstationary time series; regression; structurally changing time series; temporal structure; time series analysis; Analytical models; Computer architecture; Computer science; Feedforward neural networks; Information systems; Marine vehicles; Neural networks; Predictive models; State estimation; Time series analysis;
Conference_Titel :
Neural Networks,1997., International Conference on
Conference_Location :
Houston, TX
Print_ISBN :
0-7803-4122-8
DOI :
10.1109/ICNN.1997.614215
