![[2602.14855] A Pragmatic Method for Comparing Clusterings with Overlaps and Outliers](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.14855] A Pragmatic Method for Comparing Clusterings with Overlaps and Outliers
Summary
This paper presents a new method for comparing clustering results that accommodates overlaps and outliers, addressing a gap in existing evaluation techniques.
Why It Matters
Clustering is vital in data science, yet current methods fail to effectively compare clusterings with overlaps and outliers. This research offers a pragmatic solution, enhancing the evaluation of clustering algorithms and potentially improving their application across various domains.
Key Takeaways
- Introduces a pragmatic similarity measure for clustering comparison.
- Addresses the limitations of existing clustering evaluation methods.
- Demonstrates desirable properties and reduced biases in the new method.
- Enhances the accuracy of clustering assessments in data science.
- Potentially applicable across various fields that utilize clustering.
Computer Science > Machine Learning arXiv:2602.14855 (cs) [Submitted on 16 Feb 2026] Title:A Pragmatic Method for Comparing Clusterings with Overlaps and Outliers Authors:Ryan DeWolfe, Paweł Prałat, François Théberge View a PDF of the paper titled A Pragmatic Method for Comparing Clusterings with Overlaps and Outliers, by Ryan DeWolfe and 2 other authors View PDF HTML (experimental) Abstract:Clustering algorithms are an essential part of the unsupervised data science ecosystem, and extrinsic evaluation of clustering algorithms requires a method for comparing the detected clustering to a ground truth clustering. In a general setting, the detected and ground truth clusterings may have outliers (objects belonging to no cluster), overlapping clusters (objects may belong to more than one cluster), or both, but methods for comparing these clusterings are currently undeveloped. In this note, we define a pragmatic similarity measure for comparing clusterings with overlaps and outliers, show that it has several desirable properties, and experimentally confirm that it is not subject to several common biases afflicting other clustering comparison measures. Comments: Subjects: Machine Learning (cs.LG); Social and Information Networks (cs.SI); Combinatorics (math.CO) Cite as: arXiv:2602.14855 [cs.LG] (or arXiv:2602.14855v1 [cs.LG] for this version) https://doi.org/10.48550/arXiv.2602.14855 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Submission history F...