Communications in Information and Systems

Volume 5 (2005)

Number 1

Clustering time series, subspace identification and cepstral distances

Pages: 69 – 96

DOI: http://dx.doi.org/10.4310/CIS.2005.v5.n1.a3

Authors

Jeroen Boets (Department of Electrical Engineering (ESAT-SCD), K.U. Leuven, Leuven, Belgium)

K. de Cock (Department of Electrical Engineering (ESAT-SCD), K.U. Leuven, Leuven, Belgium)

Espinoza M. (Department of Electrical Engineering (ESAT-SCD), K.U. Leuven, Leuven, Belgium)

B. de Moor (Department of Electrical Engineering (ESAT-SCD), K.U. Leuven, Leuven, Belgium)

Abstract

In this paper a methodology to cluster time series based on measurement data is described. In particular, we propose a distance for stochastic models based on the concept of subspace angles within a model and between two models. This distance is used to obtain a clustering over the set of time series. We show how it is related to the mutual information of the past and the future output processes, and to a previously defined cepstral distance. Finally, the methodology is applied to the clustering of time series of power consumption within the Belgian electricity grid.

Keywords

clustering, time series, linear models, principal angles, canonical correlations, cepstrum, mutual information

Full Text (PDF format)