Int-EM-CEM algorithm for imprecise data. Comparison with the CEM algorithm using Monte Carlo simulations

This paper addresses the problem of fitting mixture model based-clustering to imprecise data using the CEM algorithm. Imprecise data are modelled by multivariate uncertainty zones, which constitute a generalization of multivariate interval-valued data. To estimate simultaneously the mixture model parameters and the partition from uncertainty zone data, we propose an adapted version of the CEM algorithm. Results on simulated data compare the proposed algorithm with the classical one (applied to the raw data then to the uncertain data).