基于贡献度的多标签传播重叠社区发现研究  

Multi-label Propagation for Overlapping Community Detection Based on Contribution Degree

作　　者：吴小兰[1,2] 章成志[1,3] 

机构地区：[1]南京理工大学信息管理系,南京210094 [2]安徽财经大学管理科学与工程学院,蚌埠233030 [3]江苏省数据工程与知识服务重点实验室(南京大学),南京210093

出　　处：《情报学报》2015年第9期949-957,共9页Journal of the China Society for Scientific and Technical Information

基　　金：国家社会科学基金项目“在线社交网络中基于用户的知识组织模式研究”(No.14BTQ033)、国家社会科学基金重大项目“面向突发事件应急决策的快速响应情报体系研究”(No.13&ZD174)、国家社会科学基金重点项目“大数据环境下社会舆情与决策支持方法体系研究”(No.14AZD084)资助

摘　　要：在真实世界网络中，一个人可以属于多个兴趣小组，一个蛋白质可以属于多个蛋白复合体，因此发现网络中的重叠社团结构更能准确地反映网络中真实信息。与现有一些发现重叠社区结构的算法相比，多标签传播算法具有时间复杂度低的优点，但当节点含有多个邻居社区且属于这些邻居社区的隶属度相等且都要小于设定的阈值参数时，多标签传播算法随机地从邻居节点标签中选择社区，这严重影响了算法稳定性。为此，本文提出了一种基于贡献度改进的多标签传播算法。在真实基准网和计算机生成网的测试结果表明该标签传播算法具有较好的社区发现性能，我们将该算法应用在科学网博客中“图书馆、情报与文献学”领域用户的好友关系网上，能有效地发现该领域中存在的重叠社区结构。In real-world networks, a man can belong to many different interest groups, a protein can belong to different complexes, so detecting overlapping communities is very important to achieve accurate structures about real-world networks. Compared with existing overlapping community detection algorithms, multi-label propagation algorithm has low algorithm complexity. However, when a node has muhiple belonging neighbor communities and its belonging coeffcients are equal and less than the threshold parameter, that muhi-label propagation algorithm randomly chooses one from these neighbor communities seriously affects the stablity of the multi-label propagation algorithm. Hence, we put forward an optimization of multi-label propagation algorithm based on contribution degree in this paper. The tests on real-world networks and synthetic networks show that our algorithm has stronger validity and stability, then our algorithm is applied at the social network of science bloggers in ＂library,＂ information and bibliography＂ field , and it detects the overlapping communities in these social network effectively.

关 键 词：社区发现 多标签传播算法 贡献度 社区结构 

分 类 号：G350[文化科学—情报学]