基于粒子群优化建模的贝叶斯网络结构学习方法  

A BAYESIAN NETWORKS STRUCTURE LEARNING METHOD BASED ON PARTICLE SWARM OPTIMISATION MODELLING

作　　者：李东灵[1] 

机构地区：[1]商丘职业技术学院,河南商丘476000

出　　处：《计算机应用与软件》2014年第11期178-182,共5页Computer Applications and Software

摘　　要：贝叶斯网络(BN)在不确定性的条件下表示信息和推理论证具有良好的性能,但由于其结构搜索空间的复杂性,通常将从一个数据集合中学习贝叶斯网络的结构认为是一种NP-hard的问题。基于此,提出一种新的基于粒子群优化算法建模的贝叶斯网络结构学习方法。为了学习一个贝叶斯网络的结构,该方法先使用粒子群优化算法在排序空间中进行搜索,然后运行K2算法计算每个排序的吻合度。每个排序都会有一个网络结构与之一致,该方法会返回这个网络的计分。仿真结果表明,在不同规模的数据集中,该算法相对于其他贝叶斯网络结构学习算法对不同类型的网络都具有更好的网络稳定性。Bayesian networks（ BNs） have good performances when used for representing information and reasoning under uncertain circumstances,but due to the complexity in its structure search space,a structure learning BN from a dataset is in general regarded as an NP-hard problem.Based on this,we propose a novel BN structure learning approach which is based on particle swarm optimisation（ PSO）modelling.In order to learn a structure of Bayesian network,the approach first uses PSO to search in sorting space,and then calculates the fitness of each ordering by running the K2 algorithm.Every ordering will have a corresponding network structure,and the approach will feed back the score of this network.Simulation results show that the approach can produce better network stabilities on the networks with different types compared with other BN structure learning algorithms in databases with different scales.

关 键 词：贝叶斯网络 粒子群优化 K2算法 结构学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]