Title: Bayesian network structure learning based on modified particle swarm optimisation

Authors: Lei Yang; Jue Wu; Feng Liu

Addresses: College of Computer Science and Technology, Southwest University of Science and Technology, Mianyang, Sichuan, China ' College of Computer Science and Technology, Southwest University of Science and Technology, Mianyang, Sichuan, China ' International School of Software, Wuhan University, Wuhan, Hubei, China

Abstract: Bayesian network structure learning is very important in the Bayesian network learning. The algorithm of Bayesian network structure learning based on the modified particle swarm optimisation is proposed in this paper. It is based on the analysis of the particle swarm optimisation and the feature of the Bayesian network structure. The function BIC is used as the evaluation standard of the Bayesian network structure in the algorithm. The most optimal particle is saved during the process. The mutation operation is used to reduce the possibility of local optimal solution. The scheme proposed in this paper is proved to be convergent in theory. In the simulation, the scheme is compared with the algorithm K2, and the experiment result shows that there is only one reversed edge in the Bayesian network structure obtained by the modified particle swarm optimisation compared with the standard Bayesian network structure of the Asia network. The missing and redundant edges in the Bayesian structure obtained by the MPSO are also less then missing and redundant edges in the Bayesian structure obtained by K2. We can draw a conclusion that the MPSO is better than K2, and the modified particle swarm optimisation is feasible for the Bayesian network structure learning.

Keywords: Bayesian networks; particle swarm optimisation; structure learning; modified PSO; Bayesian network learning; simulation; missing edges; redundant edges.

DOI: 10.1504/IJICT.2016.073641

International Journal of Information and Communication Technology, 2016 Vol.8 No.1, pp.112 - 121

Available online: 15 Dec 2015 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article