The TCLUST Approach to Robust Cluster Analysis

Abstract

A new method for performing robust clustering is proposed. The method is designed with the aim of ¯tting clusters with di®erent scat- ters and weights. A proportion ® of contaminating data points is also allowed. Restrictions on the ratio between the maximum and the min- imum eigenvalues of the groups scatter matrices are introduced. These restrictions make the problem to be well-de¯ned guaranteeing the ex- istence and the consistency of the sample estimators to the population parameters.