[en] The TCLUST procedure is a robust clustering procedure that performs clus-
tering with the aim of tting clusters with di erent scatters and weights. As
the corresponding objective function can be unbounded, a restriction is added
on the eigenvalues-ratio of the scatter matrices. The robustness of the method
is guaranteed by allowing the trimming of a given proportion of observations.
The resistance to contamination of that procedure will be studied. Results con-
cerning breakdown points and some new criteria in robust cluster analysis, such
as the dissolution point and the isolation robustness, will be presented.
Disciplines :
Mathematics
Author, co-author :
Ruwet, Christel ; Université de Liège - ULiège > Département de mathématique > Statistique mathématique
García-Escudero, Luis Angel; Universidad de Valladolid - UVa > Departamento de Estadística e Investigación operativa
Gordaliza, Alfonso; Universidad de Valladolid - UVa Departamento de Estadística e Investigación operativa > Departamento de Estadística e Investigación operativa
Mayo-Iscar, Agustin; Universidad de Valladolid - UVa > Departamento de Estadística e Investigación operativa
Language :
English
Title :
Impact of contamination on the TCLUST procedure
Publication date :
18 December 2011
Event name :
4th International Conference of ERCIM working group on Computing and Statistics