Robust clustering algorithms based on finite mixtures of multivariate t distribution

Providing protection against outlier in clustering data is a difficult problem. We proposed two robust clustering algorithms which integrate two modified versions of EM algorithm for mixtures t model with a model selection criterion respectively. The proposed methods can select the number of clusters component automatically by a combined component annihilation strategy and can also avoid the drawbacks of traditional mixture-based clustering algorithms - highly dependent on initialization and may converge to the boundary of the parameter space [7]. Experiment results show the contrast among different algorithms and demonstrate the effectiveness of our algorithms.