A new shape-based clustering algorithm for time series

• Published in: Information sciences 2022-09, Vol.609, p.411-428
• Main Authors: Li, Yucheng, Shen, Derong, Nie, Tiezheng, Kou, Yue
• Format: Article
• Language: English
• Subjects: Fractional correlation; Shape shift deviating; Time series clustering
• ISSN: 0020-0255; 1872-6291
• DOI: 10.1016/j.ins.2022.07.105

•We propose to use fractional order correlation to measure the relationship between two sequences and apply the normalized form of the results to create the fractional order shape-based distance between two sequences.•We use the average sequence calculation method DBA to determine the cluster center and then combine it with our distance to cluster the time series.•We also combine our distance with the center determination strategy of other clustering algorithms to execute comparative experiments.•Experiments show that our method has improved the clustering accuracy. Our proposed distance can achieve better results when combined with a variety of strategies. At the same time, in the shape-based clustering algorithm, compared with the best KShape algorithm, we can also achieve better results. Time series clustering is a research hotspot in data mining. Most of the existing clustering algorithms combine with the classical distance measure which ignore the offset of sequence shape. As a result, shape-based clustering algorithms are becoming increasingly popular. On the majority of data sets, the most representative shape-based clustering algorithm, KShape, which defines a shape-based distance with shift invariance, has been shown to outperform other algorithms. In this paper, we propose a new shape-based clustering algorithm named Fractional Order Shape-based k-cluster(FrOKShape), which defines a multi-variable shape-based distance by normalized fractional order cross-correlation and uses the DTW Barycenter Averaging (DBA) as a center computation strategy. Our distance exhibits excellent shape shift deviating properties and good compatibility integrated with a variety of existing clustering center strategies so that it can provide more potential good results. Experiments show that combining our distance with a traditional clustering algorithm produces excellent clustering indicators. In a series of comparative experiments, FrOKShape also exhibits a comparable result to the existing better shape-based clustering algorithm KShape.