TSX-Means: An Optimal K Search Approach for Time Series Clustering
Type | Proceedings paper | ||||||||
---|---|---|---|---|---|---|---|---|---|
Date | 2021 | ||||||||
Language | English | ||||||||
Author(s) | Tokotoko Jannai1, Selmaoui-Folcher Nazha1, Govan Rodrigue1, Lemonnier Hugues2 | ||||||||
Affiliation(s) | 1 : ISEA, University of New Caledonia, Nouméa, New Caledonia 2 : UMR ENTROPIE, IFREMER, Nouméa, New Caledonia |
||||||||
Meeting | DEXA 2021 - 32nd International Conference, DEXA 2021, Virtual Event, September 27–30, 2021 | ||||||||
Source | Strauss C., Kotsis G., Tjoa A.M., Khalil I. (eds) Database and Expert Systems Applications. 32nd International Conference, DEXA 2021, Virtual Event, September 27–30, 2021, Proceedings, Part II. Print ISBN 978-3-030-86474-3 Online ISBN 978-3-030-86475-0 . Lecture Notes in Computer Science, vol 12924, pp.232-238 | ||||||||
DOI | 10.1007/978-3-030-86475-0_23 | ||||||||
Abstract | Proliferation of temporal data in many domains has generated considerable interest in the analysis and use of time series. In that context, clustering is one of the most popular data mining methods. Whilst time series clustering algorithms generally succeed in capturing differences in shapes, they most often fail to perform clustering based on both shape and amplitude dissimilarities. In this paper, we propose a new time series clustering method that automatically determines an optimal number of clusters. Cluster refinement is based on a new dispersion criterion applied to distances between time series and their representative within a cluster. That dispersion measure allows for considering both shape and amplitude of time series. We test our method on datasets and compare results with those from K-means time series (TSK-means) and K-shape methods. |
||||||||
Full Text |
|