Structure learning of Bayesian network based on firefly hybrid optimization algorithm

Abstract

Aiming at the low learning efficiency and easy to fall into local optimization of Bayesian network structure learning algorithm, a hybrid optimization algorithm of fireflies (MIC-FA) is proposed. Firstly, the undirected graph containing most correct edges is obtained by the maximum information coefficient, so as to generate the initial population of MIC-FA algorithm. Then, the position of firefly individual is updated by directional movement, random movement and mutation operation in firefly optimization algorithm. Finally, the firefly with the highest Bayesian information criterion score is the best Bayesian network structure. Experimental results show that, compared with the contrast algorithm, the accuracy and optimization efficiency of the proposed algorithm are improved.