一种融合贝叶斯概率的社区结构发现方法研究  

Research on a Community Structure Detection Method Based on Bayesian Probability

作　　者：王刚[1] WANG Gang(Ankang University,Ankang 725000,China)

机构地区：[1]安康学院,陕西安康725000

出　　处：《计算机技术与发展》2019年第1期110-113,共4页Computer Technology and Development

基　　金：国家自然科学基金(61152003);陕西省社科基金(2015M004)

摘　　要：社区结构通常具有动态、不对称、模糊的特性。为了更好地发现社区结构以及描述社区成员之间的关系,针对当前方法的一些不足,对利用贝叶斯概率来改进社区结构发现的方法进行研究。贝叶斯概率在描述成员之间动态、因果、模糊关系时具有优势,通过引入信息熵,提出了一种融合贝叶斯概率的社区发现方法。该方法首先计算成员之间的贝叶斯概率,研究贝叶斯关系网络构建方法,得出成员之间不对称贝叶斯概率矩阵;然后根据系统内信息的熵相对稳定的性质,把成员间贝叶斯概率作为信息熵的概率输入,计算出新成员加入后信息熵的变化值,根据熵值变化情况来确定成员是否属于社区,从而在发现社区结构的同时,也能描述社区成员之间的不对称、动态和模糊关系。实验结果证明了该方法的有效性。Community structure is commonly dynamical,fuzzy and asymmetric.In order to find the community structure and describe the relationship between members,for the shortcomings of current methods,we try to study a new method of community detection based on Bayesian probability which has advantages in describing dynamic,causal and fuzzy relations among members.By introducing information entropy,we propose a community discovery method integrating Bayesian probability.It first calculates the Bayesian probability among members,studies the Bayesian network construction method,and obtains the asymmetric Bayesian probability matrix between members.Then according to the nature of relatively stable information entropy in the system,the Bayesian probability between members as probability input of information entropy,the variable of information entropy after new members joining is calculated to determine whether members belong to the community.Thus,the asymmetrical,dynamic and fuzzy relationships among community members can be described while the community structure is discovered.Experiment shows that the method is effective.

关 键 词：社区发现 贝叶斯概率 信息熵 数据挖掘 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]