A bootstrap-based aggregate classifier for model-based clustering (Q626243)

A clustering technique is discussed based on the finite mixture (FM) model. An EM algorithm is described for parameter estimation and Bayesian classification. A bagging (bootstrap aggregating) procedure is considered for improvement of the obtained clustering. Methods are presented for the label-switching problem solution in aggregation with unknown labels. Comparisons of the proposed procedures performance is made by simulation studies for clustering under the latent class model for nominal response variables. Results are presented of a real life sociological data analysis.