|
Activity Number:
|
345
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #305906 |
|
Title:
|
An Interpolated Periodogram-Based Metric for Comparison of Time Series with Unequal Lengths
|
|
Author(s):
|
Jorge Caiado*+ and Nuno Crato and Daniel Peña
|
|
Companies:
|
CEMAPRE/ISEG and IPS and CEMAPRE/Technical University of Lisbon and Universidad Carlos III de Madrid
|
|
Address:
|
Rua do Quelhas 6, Lisboa, 1200-781, Portugal
|
|
Keywords:
|
classification ; ARMA and ARFIMA models ; time series ; periodogram ; cluster analysis
|
|
Abstract:
|
The comparison of time series has been studied in literature using both time and frequency domain methods. However, existing spectral methods for discrimination and clustering analysis of time series cannot be applied directly to series with unequal lengths. Some studies use time series of unequal length and had to truncate time series spectra to compare them. We then try and develop a method without this drawback. We propose a periodogram-based method for classifying times series with different lengths. For such cases, we know that the Euclidean distance between the periodogram ordinates cannot be used. One possible way to deal with this problem is to interpolate linearly one of the periodograms in order to estimate ordinates of the same frequencies. The interpolated periodogram-based is used to compare different stationary and near-stationary processes.
|
