Advanced Analytics and Learning on Temporal Data - 8th ECML PKDD Workshop, AALTD 2023, Turin, Italy, September 18-22, 2023, Revised Selected Papers
Advanced Analytics and Learning on Temporal Data - 8th ECML PKDD Workshop, AALTD 2023, Turin, Italy, September 18-22, 2023, Revised Selected Papers
复制标题
时态数据的高级分析和学习 - 第 8 届 ECML PKDD 研讨会，AALTD 2023，意大利都灵，2023 年 9 月 18-22 日，修订后的精选论文
DOI：
10.1007/978-3-031-49896-1_4
复制
发表时间：
2023
期刊：
影响因子：
--
通讯作者：
Holder C
中科院分区：
文献类型：
--
作者：
Holder C
时间序列聚类（TSCL）涉及将未标记的时间序列分组为同质组。一种流行的TSCL方法是使用分区聚类算法sk-means ork-medoids结合弹性距离函数，如动态时间弯曲（DTW）。我们使用九种不同的弹性距离测量来探索TSCL。这两种分区算法都使用样本序列来聚类，但使用不同的技术来实现：k-means使用平均算法来找到样本，而medoids选择训练案例（medoid）。传统上，时间序列集合的算术平均值与k-means一起使用。但是，这会忽略任何偏移。2011年，提出了一种特定于DTW的平均技术，称为DTW重心平均（DBA）。从那时起，k-means和DBA一直是大多数基于分区的TSCL的首选算法，并且许多使用基于medoids的TSCL方法的研究已经停止。我们修订k-medoids基于TSCL与一系列的弹性距离措施。我们的研究结果表明，k-medoids方法在标准测试套件上显着更好的感谢手段，独立于所使用的弹性距离测量。我们还比较了最常用的交替k-medoids方法对周围的Medoids（PAM）算法的分区。PAM显着优于默认的k-medoids的所有九个弹性措施。此外，我们评估了六种PAM变体，旨在加快TSCL。最后，我们证明了具有最佳弹性距离度量的PAM明显优于流行的替代TSCL算法，包括k-means DBA方法，并与最好的深度学习算法竞争。
Time Series Clustering (TSCL) involves grouping unlabelled time series into homogeneous groups. A popular approach to TSCL is to use the partitional clustering algorithmsk-means ork-medoids in conjunction with an elastic distance function such as Dynamic Time Warping (DTW). We explore TSCL using nine different elastic distance measures. Both partitional algorithms characterise clusters with an exemplar series, but use different techniques to do so:k-means uses an averaging algorithm to find an exemplar, whereask-medoids chooses a training case (medoid). Traditionally, the arithmetic mean of a collection of time series was used withk-means. However, this ignores any offset. In 2011, an averaging technique specific to DTW, called DTW Barycentre Averaging (DBA), was proposed. Since,k-means with DBA has been the algorithm of choice for the majority of partition-based TSCL and much of the research using medoids-based approaches for TSCL stopped. We revisitk-medoids based TSCL with a range of elastic distance measures. Our results showk-medoids approaches are significantly better thank-means on a standard test suite, independent of the elastic distance measure used. We also compare the most commonly used alternatingk-medoids approach against the Partition Around Medoids (PAM) algorithm. PAM significantly outperforms the defaultk-medoids for all nine elastic measures used. Additionally, we evaluate six variants of PAM designed to speed up TSCL. Finally, we show PAM with the best elastic distance measure is significantly better than popular alternative TSCL algorithms, including thek-means DBA approach, and competitive with the best deep learning algorithms.