基于交互行为的在线社会网络社区发现方法研究  

New Method of Community Detection for Online Social Networks Based on Interactive Behaviors

作　　者：田博[1] 凡玲玲 

机构地区：[1]上海财经大学信息管理与工程学院,上海200433

出　　处：《情报杂志》2016年第11期183-188,共6页Journal of Intelligence

基　　金：国家自然科学基金项目"Bto C电子商务中基于分层Bayesian网络的信任与声誉计算理论研究"(编号:71302080);教育部人文社科基金"移动支付中的感知信任与消费决策理论研究"(编号:13YJC630149)资助

摘　　要：[目的/意义]社区发现是社交网络、生物基因表达以及物理系统等领域的一个重要且具有挑战性的研究课题。现有的在线社会网络社区发现方法可分为基于静态关系的社区发现和基于动态关系的社区发现。相对于静态社区发现方法,动态交互行为关系更能体现社交媒体用户的关系紧密程度。[方法/过程]提出了一种基于交互行为的在线社会网络社区发现方法,利用用户之间的转发关系、评论关系、点赞关系以及提及关系等四方面交互行为构建加权的交互网络;然后以Newman模块度函数作为优化目标,得到社区内部节点联系紧密、社区之间节点联系松散的社区划分方法。具体实现上以改进模块度函数为指标,不断合并模块度函数增益最大的节点对,直至所有的模块度函数增益都为负。[结果/结论]提出的方法在如下两方面进行改进:一方面建立了交互行为公式,同时对模块度函数进行改进。最后,选取新浪微博数据对提出的算法进行实验验证,实验结果表明提出的基于交互行为社区发现方法的有效性。[ Purpose/Significance] Community detection is an important and challenging problem in several research fields such as social complex networks, biology gene expression, and physics systems and so on. There are two popular types of community detection for the online social network in general nowadays. The first one is based on the statistic relationships between nodes. The second one is based on the dynamical relationships between nodes. The dynamical interactive relationship can better reflect the real relationship between social media users. [ Method/Proceas ] A new method of community detection for online social networks is proposed based on the interactive behavior of the nodes of networks in this paper. Based on the four aspects of the interaction between users such as sharing, commenting, agreeing and mentioning relationship, a weighted interactive network is constructed. Those nodes that have the maximum and positive module gain are merged until all the module gains are negative. [ Resnit/Conclusion ] Compared to exiting methods, the method has two merits such as the module function is improved and interactions of nodes are measured by formula. At last, experiments on the data from Sina Weibo verify the proposed method and show its effectiveness.

关 键 词：在线社交网络 社区发现 交互行为 模块度函数 动态网络 模块度增益 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]