End-to-end deep representation learning for time series clustering: a comparative study
End-to-end deep representation learning for time series clustering: a comparative study
复制标题
DOI：
10.1007/s10618-021-00796-y
复制
发表时间：
2021-10-16
影响因子：
4.8
通讯作者：
Forestier, Germain
中科院分区：
文献类型：
--
作者：
Lafabregue, Baptiste;Weber, Jonathan;Forestier, Germain
时间序列在数据挖掘应用中无处不在。与其他类型的数据类似，注释可能很难获取，从而妨碍训练时间序列分类模型。在这种情况下，聚类方法可能是一个合适的替代方法，因为它们创建同质组，从而可以更好地分析数据结构。时间序列聚类已经研究了很多年，并且已经提出了多种方法。随着计算机视觉深度学习的出现，研究人员最近开始研究使用深度聚类对时间序列数据进行聚类。现有的方法主要依赖于表示学习（从计算机视觉导入），其中包括学习数据的表示并使用这种新表示执行聚类任务。本文的目的是对深度聚类时间序列表示学习的现有文献进行仔细研究和实验比较。在本文中，我们超越了现有方法的唯一比较，并提出将深度聚类方法分解为三个主要组成部分：（1）网络架构，（2）借口损失和（3）聚类损失。我们评估了这些组件的所有组合（总共 300 个不同的模型），目的是研究它们对聚类性能的相对影响。我们还通过实验比较了我们确定的最有效的组合与现有的非深度聚类方法。使用由 128 个单变量数据集和 30 个多元数据集组成的最大时间序列数据集存储库（UCR/UEA 存档）进行实验。最后，我们提出了将类激活映射方法扩展到无监督情况，该方法允许识别模式，从而突出显示网络如何对时间序列进行聚类。
Time series are ubiquitous in data mining applications. Similar to other types of data, annotations can be challenging to acquire, thus preventing from training time series classification models. In this context, clustering methods can be an appropriate alternative as they create homogeneous groups allowing a better analysis of the data structure. Time series clustering has been investigated for many years and multiple approaches have already been proposed. Following the advent of deep learning in computer vision, researchers recently started to study the use of deep clustering to cluster time series data. The existing approaches mostly rely on representation learning (imported from computer vision), which consists of learning a representation of the data and performing the clustering task using this new representation. The goal of this paper is to provide a careful study and an experimental comparison of the existing literature on time series representation learning for deep clustering. In this paper, we went beyond the sole comparison of existing approaches and proposed to decompose deep clustering methods into three main components: (1) network architecture, (2) pretext loss, and (3) clustering loss. We evaluated all combinations of these components (totaling 300 different models) with the objective to study their relative influence on the clustering performance. We also experimentally compared the most efficient combinations we identified with existing non-deep clustering methods. Experiments were performed using the largest repository of time series datasets (the UCR/UEA archive) composed of 128 univariate and 30 multivariate datasets. Finally, we proposed an extension of the class activation maps method to the unsupervised case which allows to identify patterns providing highlights on how the network clustered the time series.