Mixture Model-based Clustering Algorithm Research

Cluster analysisi is the problem of determining the intrinsic structure of clustered data when no information other than the observed values is available. Probability models have been proposed for quite some time as a basis for cluster analysis. Finite mixture models are an increasingly important tool in multivariate statistics. Approaches to density estimation and clustering based on normal mixture models have shown good performance in many applications. In this approach, the data are viewed as coming from a mixture of probability distributions, each representing a different cluster. We describe a clustering methodology based on multivariate normal mixtures in which the MLE is replaced by a maximum a posteriori (MAP) estimator that may avoid singularity.This paper is organized as follows. In Section 1,we give the necessary background in cluster analysis. In Section 2,we review finite mixture models, the EM algorithm. In Section 3, we give a brief overview of model-based clustering, which can also be viewed as a proceedure for normal mixture estimation that includes model selection, both in terms of component structure and number of components. In Section 4,we proposed an unsupervised clustering algothrim based on MAP estimator, not only avoid singularity, but also give a good performance on model selection.