WSEAS Transactions on Computers
Print ISSN: 1109-2750, E-ISSN: 2224-2872
Volume 13, 2014
Change-Point Detection in Multivariate Time-Series Data by Recurrence Plot
Authors: , , , ,
Abstract: Change-point detection in time-series is an important data mining task with applications to abnormity diagnosis, events monitoring, climate change analysis, and other domains. This paper presents a novel method based on recurrence plot for detecting multiple change-points in multivariate time series. Bhattacharyya distance function is applied to improve the recurrence plot generation so as to capture the dependency change among variables. A window-based detection algorithm is proposed to capture the change-points quickly and automatically. With experiments on artificial and real datasets, we show that the algorithm has made improvement to traditional recurrence plot, is able to handle noisy data with optimized parameter, and can cope with complex situation like human activity and micro-blog events monitoring.
Search Articles
Pages: 592-599
WSEAS Transactions on Computers, ISSN / E-ISSN: 1109-2750 / 2224-2872, Volume 13, 2014, Art. #53