Since it is an NP-hard problem to learn a Bayesian network from data, when the number of data is large, the process of learning Bayesian networks structure is prone to get into premature convergence, then obtain local...
详细信息
Since it is an NP-hard problem to learn a Bayesian network from data, when the number of data is large, the process of learning Bayesian networks structure is prone to get into premature convergence, then obtain local optimal Bayesian networks under particleswarmoptimizationalgorithm(PSO), a novel approach is proposed. The binary code quantum-behaved particle swarm optimization algorithm(BCQPSO) and Bayesian Information Criterion score(BIC) are applied to obtain an optimal Bayesian network. In order to evaluate the matching degree between Bayesian networks and original network structure, BIC score is proposed. The proposed algorithm–BCQPSO is to seek out Bayesian networks that match sample data sets and have a higher BIC score. Besides, in order to get a more optimal network structure, the process of searching has deleted cycle. The ASIA network is used to test the proposed technique, which has a better performance than K2, Maximum Weight Spanning Tree(MWST), and PSO.
暂无评论