An improved Bayesian structural EM algorithm for learning Bayesian networks for clustering

The application of the Bayesian Structural EM algorithm to learn Bayesian n etworks (BNs) for clustering implies a search over the space of BN structur es alternating between two steps: an optimization of the BN parameters (usu ally by means of the EM algorithm) and a structural search for model select ion. In this paper, we propose to perform the optimization of the BN parame ters using an alternative approach to the EM algorithm: the BC + EM method. We provide experimental results to show that our proposal results in a mor e effective and efficient version of the Bayesian Structural EM algorithm f or learning BNs for clustering. (C) 2000 Elsevier Science B.V. All rights r eserved.