The MBC Procedure

The MBC procedure fits mixtures of multivariate Gaussian and uniform distributions to achieve unsupervised and semisupervised clustering of data. It treats the cluster memberships as missing data and uses the expectation-maximization (EM) algorithm to maximize the likelihood. The procedure can produce output data tables that contain cluster membership weights and component log likelihoods. PROC MBC can also store the fitted model for scoring subsequent input data.

PROC MBC assumes that the data can be described by a mixture of multivariate Gaussian components and an optional noise component. The procedure also assumes that the data are continuous, and it does not support CLASS variables.