Clusters of Multivariate Stationary Time Series by Differential Evolution and Autoregressive Distance
Clustering MTS is a difficult task that has to be performed in several application fields. We propose a method based on the coefficients of vector autoregressive (VAR) models and differential evolution (DE) that may be applied to sets of stationary MTS. Results from a simulation experiment that includes both linear and non linear MTS are displayed for comparison with genetic algorithms (GAs), principal component analysis (PCA) and the k-means algorithm. Part of the Australian Sign Language (Auslan) data are examined to show the comparative performance of our procedure on a real world data set.
KeywordsAutoregressive distance Cluster analysis Differential evolution Multivariate time series
- 1.Allefeld, C., Bialonski, S.: Detecting syncronization clusters in multivariate time series via coarse-graining of Markov chains. Tech. rep., arXiv:0707.2479v1 [physics.data-an] (July 17, 2007)Google Scholar
- 2.Bandyopadhyay, S., Baragona, R., Maulik, U.: Clustering multivariate time series by genetic multiobjective optimization. Metron 68, 161–183 (2010)Google Scholar
- 17.Yang, K., Shahabi, C.: A PCA-based similarity measure for multivariate time series. In: MMDB 2004: Proceedings of the 2nd ACM international workshop on Multimedia databases, pp. 65–74. ACM, New York (2004)Google Scholar