Clustering with Mixed Type Variables and Determination of Cluster Numbers

Authors: H. Řezanková, Dušan Húsek, T. Löster
Publication year: 2010
Publication type: Konferenční příspěvek (mezinárodní konf.)
Periodical: Proceedings of COMPSTAT 2010
Projects: Decompositions of matrices with binary and ordinal data: theory, algorithms, and complexity
Methods of Artificial Inteligence in GIS
Bibliographic citation: ŘEZANKOVÁ, H., HÚSEK, Dušan, LÖSTER, T. Clustering with Mixed Type Variables and Determination of Cluster Numbers. In: LECHEVALLIER, Y, SAPORTA, G., eds. Proceedings of COMPSTAT 2010. Heidelberg: Physica Verlag, 2010, S. 1525-1532. ISBN 978-3-7908-2603-6.
ID: 347817

Abstract

The main applications of cluster analysis methods concern clustering objects characterized by quantitative variables. In the paper, the evaluation criteria for the case of mixed type variables are dealt with. The criteria based on variability measures for nominal and quantitative variables are described. A combination of variance and entropy or Gini's coefficient of mutability is used. The coefficients for determination of cluster numbers based on variability measures for mixed type variables are applied to a real-data to show viability of our approach.