LEARNING BAYESIAN NETWORK STRUCTURES BY SEARCHING FOR THE BEST ORDERING WITH GENETIC ALGORITHMS

In this paper we present a new methodology for inducing Bayesian netwo rk structures from a database of cases. The methodology is based on se arching for the best ordering of the system variables by means of gene tic algorithms. Since this problem of ending an optimal ordering of va riables resembles the traveling salesman problem, we use genetic opera tors that were developed for the latter problem. The quality of a vari able ordering is evaluated with the structure-learning algorithm K2. W e present empirical results that were obtained with a simulation of th e ALARM network.