社会网络顶点间相似性度量及其应用

发布时间：2018-04-06 07:34

本文选题：社会网络　切入点：相似性　出处：《计算机科学与探索》2017年10期

【摘要】：集对分析作为处理系统确定性与不确定性相互作用的数学理论,可用来处理存在不确定关系的复杂社会网络。首先,应用集对分析理论,将社会网络作为一个同异反系统(确定不确定系统),采用集对联系度刻画顶点间的同异反关系,综合考虑顶点的局部特征和拓扑结构对顶点相似性的贡献,提出加权聚集系数联系度的顶点间相似性度量方法。该度量方法可以更好地刻画网络结构特征,克服传统局部相似性度量指标对某些顶点间相似性值的低估,降低全局相似性度量指标的计算复杂度。其次,为了将该相似性度量指标应用于社区发现,与凝聚型层次聚类算法相结合,使其适用于具有相似性度量对象的复杂网络社区发现问题。最后,在社会网络上进行社区挖掘实验,并与经典社区发现算法进行比较,实验结果表明了该相似性度量指标的正确性及有效性。
[Abstract]:Set pair analysis, as a mathematical theory dealing with the interaction between certainty and uncertainty, can be used to deal with complex social networks with uncertain relationships.First of all, using set pair analysis theory, the social network is regarded as an identical, different and inverse system. Set pair connection degree is used to describe the similarity, difference and inverse relationship between vertices.Considering the local characteristics of vertices and the contribution of topological structure to vertex similarity, a method of measuring the similarity between vertices is proposed.The proposed method can better characterize the network structure, overcome the underestimation of some vertex similarity values by traditional local similarity metrics, and reduce the computational complexity of global similarity metrics.Secondly, in order to apply the similarity metric to community discovery, it is combined with the condensed hierarchical clustering algorithm to make it suitable for the complex network community discovery problem with similarity measurement object.Finally, community mining experiments are carried out on social networks and compared with the classical community discovery algorithms. The experimental results show that the similarity metrics are correct and effective.
【作者单位】：燕山大学信息科学与工程学院;华北理工大学迁安学院;河北省计算机虚拟技术与系统集成重点实验室;华北理工大学理学院;
【基金】：国家自然科学基金No.61472340 国家青年科学基金项目No.61602401 河北省自然科学基金Nos.F2017209070,F2016209344~~
【分类号】：O157.5;TP301.6

【相似文献】