Robust Model-Based Clustering

Abstract

We propose a class of Fisher-consistent robust estimators for mixture models. These estimators are then used to build a robust model-based clustering procedure. We study in detail the case of multivariate Gaussian mixtures and propose an algorithm, similar to the EM algorithm, to compute the proposed estimators and build the robust clusters. An extensive Monte Carlo simulation study shows that our proposal outperforms other robust and non robust, state of the art, model-based clustering procedures. We apply our proposal to a real data set and show that again it outperforms alternative procedures.