Robust fuzzy clustering using mixtures of Student’s-t distributions

In this paper, we propose a robust fuzzy clustering algorithm, based on a fuzzy treatment of finite mixtures of multivariate Student’s-t distributions, using the fuzzy c-means (FCM) algorithm. As we experimentally demonstrate, the proposed algorithm, by incorporating the assumptions about the probabilistic nature of the clusters being dirived into the fuzzy clustering procedure, allows for the exploitation of the hard tails of the multivariate Student’s-t distribution, to obtain a robust to outliers fuzzy clustering algorithm, offering increased clustering performance comparing to existing FCM-based algorithms. Our experimental results prove that the proposed fuzzy treatment of finite mixtures of Student’s-t distributions is more effective comparing to their statistical treatments using EM-type algorithms, while imposing comparable computational loads.