STRUCTURE LEARNING OF BAYESIAN NETWORKS BY GENETIC ALGORITHMS - A PERFORMANCE ANALYSIS OF CONTROL PARAMETERS

We present a new approach to structure learning in the field of Bayesi an networks: We tackle the problem of the search for the best Bayesian network structure, given a database of cases, using the genetic algor ithm philosophy for searching among alternative structures. We start b y assuming an ordering between the nodes of the network structures. Th is assumption is necessary to guarantee that the networks that are cre ated by the genetic algorithms are legal Bayesian network structures. Next, we release the ordering assumption by using a ''repair operator' ' which converts illegal structures into legal ones. We present empiri cal results and analyze them statistically. The best results are obtai ned with an elitist genetic algorithm that contains a local optimizer.