An examination of indexes for determining the number of clusters in binary data sets

Andreas Weingessel, Evgenia Dimitriadou, Sara Dolnicar

Publikation: Working/Discussion PaperWU Working Paper

74 Downloads (Pure)

Abstract

An examination of 14 indexes for determining the number of clusters is conducted on artificial binary data sets being generated according to various design factors. To provide a variety of clustering solutions the data sets are analyzed by different non hierarchical clustering methods. The purpose of the paper is to present the performance and the ability of an index to detect the proper number of clusters in a binary data set under various conditions and different difficulty levels. (author's abstract)

Publikationsreihe

ReiheWorking Papers SFB "Adaptive Information Systems and Modelling in Economics and Management Science"
Nummer29

WU Working Paper Reihe

  • Working Papers SFB \Adaptive Information Systems and Modelling in Economics and Management Science\

Zitat