go-v.orgdoi

Skeleton-based validation for density-based clustering

Christian Braune · Digitalen Hochschulbibliothek Sachsen-Anhalt (Universitäts- und Landesbibliothek Sachsen-Anhalt) · 2018

  • DOI 10.25673/13402
  • Publisher Universitäts Frauenklinik
  • Type other
  • Issued 2018

Not cached yet. This record was fetched live for this request and is stored after the lookup. Reopening the page should serve the cached copy.

Digitalen Hochschulbibliothek Sachsen-Anhalt (Universitäts- und Landesbibliothek Sachsen-Anhalt) · Universitäts Frauenklinik · other · 2018

Abstract

Clustering is an important process in data analysis. It is the process of grouping previously unlabeled data and distinguishing noise or outliers from interesting data. Clustering algorithms can work in many different ways. These are, for example, centroid-based methods like k-means, hierarchical clustering or densitybased methods. DBSCAN is the best-known representative of the latter. Since clustering is an unsupervised learning approach, validation measures are needed to assess whether a found result is good or not. In the case of centroid-based algorithms there exists a plethora of validation measures for the crisp and the fuzzy case.

Only recently the validation of density-based clustering has made some progress. The majority of the hitherto used validation measures refer in some way or another to the centroids of the clusters. When using density-based clustering, these centroids to not exist or have no meaning. Therefore, these measures are only help- or meaningful under a limited scope of scenarios. They would favor clusterings that—to the human eye—are obviously sub-optimal. This thesis suggest a way to make centroid-based cluster validation measures available for clusterings obtained by a density-based algorithm.

For this the arithmetic mean as centroid is replaced by a cluster skeleton that provides more structural information for a cluster than a single point. It can then be used instead of the former centroid in the calculation of the validation scores. After discussing several different techniques of finding such an object, the behavior of the different validation scores is analyzed.

Open the resolved locator · 10.25673/13402

Metadata licence: OpenAlex metadata (CC0)

Registry
OpenAlex DOI index + API
Fetched
2026-11-20 01:51 UTC
Cache state
not cached yet (bulk-index+openalex)
Locator
https://opendata.uni-halle.de//handle/1981185920/13465
Metadata licence
OpenAlex metadata (CC0)
Scope
Metadata as returned by the resolver; abstracts are never generated here.