Answers are grounded in this paper and cite their sources.
Novel algorithm CluMap finds matches and clusters in categorical time series, indicating improved accuracy over baseline methods.
Categorical time series (CTS) are time series that take their values from categorical types. Despite the fact that they have a broad range of applications, they have been less studied than other types of time series. In this paper, we study the problem of finding matches between disjoint sets of unordered categorical types thanks to CTS taking their values in them. To tackle this problem, we propose a novel iterative algorithm called Cluster and Map (CluMap) that finds matches using feature-based similarities of the elements of the above sets in their respective CTS. Our algorithm discovers these similarities across pair-wise sets to be mapped and then finds clusters of elements in one set to be associated with clusters in the opposite set. We also propose specific evaluation metrics for this problem, since classical metrics for quality assessment of clustering are not suitable. We experimentally validate our approach on both synthetic and real-world CTS showing its superiority with respect to baselines and its underlying accuracy.
No takes yet. Share an insight, caveat, or question.
Durand et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: