A new similarity measure and its use in determining the number of clusters in a multivariate data set

dc.contributor.author	Vassiliou, A	en
dc.contributor.author	Tambouratzis, DG	en
dc.contributor.author	Koutras, MV	en
dc.contributor.author	Bersimis, S	en
dc.date.accessioned	2014-06-06T06:45:53Z
dc.date.available	2014-06-06T06:45:53Z
dc.date.issued	2004	en
dc.identifier.issn	03610926	en
dc.identifier.uri	http://dx.doi.org/10.1081/STA-120037266	en
dc.identifier.uri	http://62.217.125.90/xmlui/handle/123456789/2690
dc.subject	Andrews' plots	en
dc.subject	Distance measures	en
dc.subject	Number of clusters	en
dc.subject	Similarity measure	en
dc.subject	Success runs	en
dc.subject.other	Algorithms	en
dc.subject.other	Data acquisition	en
dc.subject.other	Graph theory	en
dc.subject.other	Hierarchical systems	en
dc.subject.other	Number theory	en
dc.subject.other	Set theory	en
dc.subject.other	Andrews' plots	en
dc.subject.other	Distance measures	en
dc.subject.other	Number of clusters	en
dc.subject.other	Similarity measure	en
dc.subject.other	Success runs	en
dc.subject.other	Statistics	en
dc.title	A new similarity measure and its use in determining the number of clusters in a multivariate data set	en
heal.type	journalArticle	en
heal.identifier.primary	10.1081/STA-120037266	en
heal.publicationDate	2004	en
heal.abstract	Krolak-Schwerdt and Eckes [Krolak-Schwerdt, S., Eckes, T. (1992). A graph theoretic criterion for determining the number of cluster in a data set. Multivariate Behav. Res. 27(4):541-565] suggested a graph theoretic criterion, named GRAPH, which can be used to decide on the number of clusters present in a data set. However, the resulting algorithm usually terminates with fewer groups than are, actually, present in the data set. To alleviate this effect we first introduce a new distance measure based on success runs theory and Andrews curves [Andrews, D. F. (1972). Plots of high dimensional data. Biometrics 28:125-136] and incorporate it in the GRAPH procedure (as an alternative to the standard Euclidean distance used there). Extensive numerical experimentation revealed that the new algorithm behaves much better than the original GRAPH algorithm.	en
heal.journalName	Communications in Statistics - Theory and Methods	en
dc.identifier.issue	7	en
dc.identifier.volume	33	en
dc.identifier.doi	10.1081/STA-120037266	en
dc.identifier.spage	1643	en
dc.identifier.epage	1666	en