Source note2 min read
How to evaluate clustering with ground truth?
Source
arXiv
arxiv.org/abs/2606.27061v1
Source
- Source name: arXiv
- Source kind: paper
- Authors / organization: Pasi Fränti
- Venue: arXiv: cs.AI
- Published: 2026-06-25
- Collected: 2026-06-27
- URL: http://arxiv.org/abs/2606.27061v1
- PDF: https://arxiv.org/pdf/2606.27061v1
- DOI: UNKNOWN
Plain-Language Brief
The text discusses the use of external validation indexes for evaluating clustering algorithms when ground truth is available. It focuses on set-matching-based measures and recommends the Centroid Index (CI) as a simple, intuitive cluster-level measure. For more fine-tuned evaluations, options like Pair-set Index (PSI) and Clustering Accuracy are suggested.
What the Document Says
- External indexes can be used for clustering evaluation with ground truth.
- Set-matching-based measures are a focus.
- Centroid Index (CI) is recommended as an intuitive, cluster-level measure.
- PSI provides a normalized score and is not biased by cluster sizes.
- If all points should matter equally, set-matching measures like Clustering Accuracy are suitable.
Key Concepts Mentioned
- clustering evaluation
- ground truth
- external validity indexes
- set-matching-based measures
- Centroid Index (CI)
- Pair-set Index (PSI)
- Clustering Accuracy
Claims Made by the Document
- External indexes can be used for clustering evaluation with ground truth.
- Set-matching-based measures are useful for evaluating clustering algorithms.
- Centroid Index is an intuitive, cluster-level measure.
- PSI provides a normalized score that's not biased by cluster sizes.
- If all points should matter equally, set-matching measures like Clustering Accuracy are suitable.
Evidence Used
- The text discusses the use of external validation indexes for evaluating clustering algorithms.
- It focuses on set-matching-based measures.
- Centroid Index is recommended as an intuitive, cluster-level measure.
- PSI's normalized score and lack of bias by cluster sizes are mentioned.
- Clustering Accuracy is suggested when all points should matter equally.
Methods / Approach
UNKNOWN
Limitations Stated by the Authors
UNKNOWN
Limitations Noted by the Agent
Agent interpretation — not from the source.
UNKNOWN
Open Questions
UNKNOWN
Notes for Future Q&A
May include interpretation. Labeled where it does.
UNKNOWN
Raw Abstract / Excerpt
External indexes can be used for cluster evaluation when ground truth is available. We review the most common external validity indexes focusing on set-matching-based measures. We recommend centroid index (CI), because it is an intuitive cluster-level measure with an explainable result. If we need a more fine-tuned, point-level measure, there are more choices. Pair-set index (PSI) provides a normalized score which is not biased by cluster sizes. If all points should matter equally, then clustering accuracy (ACC) or any other set-matching measure is suitable.
