Arbitrary Shaped Clustering Validation on the Test Bench

Georg Stefan Schlake, Christian Beecks

2025

Abstract

Clustering is a highly important as well as highly subjective task in the field of data analytics. Selecting a suitable clustering method and a good clustering result is all but trivial and needs insight into not only the field of clustering, but also the application scenario, in which the clustering is utilized. Evaluating a single clustering is hard, especially as there exists a wide variety of indices to evaluate the quality of a clustering, both for simple convex and for arbitrary shaped clusterings. In this paper, we investigate the ability of 11 state-of-the-art Clustering Validation Indices (CVI) to evaluate arbitrary shaped clusterings. To this end, we provide a survey of the intuitive workings of these CVI and an extensive benchmark on newly generated datasets. Furthermore, we evaluate both the Euclidean distance and the density-based DC-distance to quantify the quality of arbitrary shaped clusters. We use the generation of novel datasets to evaluate the influence of a number of metafeatures on the CVI.

Paper Citation

in Harvard Style

Schlake G. and Beecks C. (2025). Arbitrary Shaped Clustering Validation on the Test Bench. In Proceedings of the 14th International Conference on Data Science, Technology and Applications - Volume 1: DATA; ISBN 978-989-758-758-0, SciTePress, pages 363-373. DOI: 10.5220/0013495500003967

in Bibtex Style

@conference{data25,
author={Georg Stefan Schlake and Christian Beecks},
title={Arbitrary Shaped Clustering Validation on the Test Bench},
booktitle={Proceedings of the 14th International Conference on Data Science, Technology and Applications - Volume 1: DATA},
year={2025},
pages={363-373},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013495500003967},
isbn={978-989-758-758-0},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 14th International Conference on Data Science, Technology and Applications - Volume 1: DATA
TI - Arbitrary Shaped Clustering Validation on the Test Bench
SN - 978-989-758-758-0
AU - Schlake G.
AU - Beecks C.
PY - 2025
SP - 363
EP - 373
DO - 10.5220/0013495500003967
PB - SciTePress