Skip to Main Content
Searching the best Bayesian Network is an NP-hard problem. When the number of variables in Bayesian Network is large, the process of searching is likely to fall into premature convergence and return a local optimal network structure. A new approach for Bayesian Networks structure learning, which is based on the discrete Binary Quantum-behaved Particle Swarm Optimization algorithm, is introduced. The proposed approach is used to find a Bayesian Network that best matches sample data sets. For evaluating the best matching degree between Bayesian Network and sample data sets, Bayesian Information Criterion score is proposed. Then ASIA network, a benchmarks of Bayesian Networks, is used to test the new approach. The results of experiment show that the proposed technique converges more rapidly than other evolutionary computation methods.