Impact of contamination on training and test error rates in statistical clustering

Christel Ruwet (1) , Gentiane Haesbroeck (2)

1 Mathematics
2 Mathematics

The k-means algorithm is one of the most common nonhierarchical clustering methods. However, this procedure is not robust with respect to atypical observations. Alternative techniques have thus been introduced, e.g. the generalized k-means procedure. In this paper, focus is on the error rate these procedures achieve when one expects the data to be distributed according to a mixture distribution. Two different definitions of the error rate are under consideration, depending on the data at hand.

Mots clés: Physical Sciences

Domaines: Calcul [stat.CO]

Christel Ruwet, Gentiane Haesbroeck. Impact of contamination on training and test error rates in statistical clustering. Communications in Statistics - Simulation and Computation, 2011, 40 (03), pp.394-411. ⟨10.1080/03610918.2010.542847⟩. ⟨hal-00666340⟩