Answers are grounded in this paper and cite their sources.
Time-series clustering is widely used in practice, yet the resulting clusters and assignments are often hard to interpret. Moreover, there is a lack of tools that explain both how a specific series can move across cluster boundaries and how such transitions can be summarized at the cluster level. In this paper, we present Galactica, an interactive visual system for exploring local and global counterfactual explanations in time-series clustering. The system is built on Galactic, a unified framework that generates minimal, structurally meaningful perturbations for individual series while also producing compact summaries of cluster-to-cluster transitions using the Minimum Description Length (MDL) principle. Galactica allows users to inspect the structure of a clustered time-series collection, analyze internal cluster organization, explore instance-level counterfactual explanations, examine global summaries of cluster transitions, and compare Galactic against state-of-the-art baselines. By coupling the algorithmic internals of Galactic with an interactive visual interface, Galactica turns abstract cluster transitions into interpretable and auditable explanation workflows.
No takes yet. Share an insight, caveat, or question.
Fragkathoulas et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: