摘要
介绍了贝叶斯网络的基本概念及其推理模式和推理算法,重点进行了团树传播算法研究,应用Matlab软件的jtree_inf_engine模块实现了团树传播算法编程,并以停车行为分析的贝叶斯网络为例进行了预测、诊断、原因关联和混合4种推理分析,根据推理结果总结了城市居民的停车决策行为特征。研究结果表明,团树传播算法可用于贝叶斯网络未知参数的推理和变量间的关系学习。
This paper introduces the basic concepts,inference modes and algorithm of Bayesian network.It focuses on Bayesian network＇s clique tree propagation algorithm.The programme of the clique tree propagation algorithm is realized by employing the jtree_inf_engine module of Matlab software.By taking parking behavior analysis of Bayesian network as an example,prediction inference,diagnostic inference,causal inference and hybrid inference are carried out.The characteristics of decision-making about parking are summarized according the inference results.Research results indicate that clique tree propagation algorithm can be adopted to infer the unknown variable of the Bayesian network and to examine the relationships among the variables.
出处
《长春大学学报》
2012年第5期505-509,共5页
Journal of Changchun University
基金
吉林省教育厅"十二五"重点项目(2011187)
国家自然科学基金资助项目(50908099)
博士后科学基金面上资助项目(20100481055)
关键词
贝叶斯网络
团树传播算法
停车行为
概率推理
Bayesian network
clique tree propagation algorithm
parking behavior
probabilistic inference
