基于社团结构的多粒度结构洞占据者发现及分析

摘要

Recently, more and more attentions have been paid to research of structural holes, and some methods have been proposed to identify the structural holes based on the community structure. However, the network indi⁃cates a hierarchical structure after dividing into communities in different granularity, and influences the nodes’ ex⁃tent to span structural holes in community structure. A structural hole spanners mining algorithm, named MG_MaxD, is proposed which is in a hierarchical network based on the idea of network community division. First,differ⁃ent granular communities are partitioned by using hierarchical community dividing algorithm ( such as EAGLE in this paper) . Then, structural hole spanners mining algorithm MG_MaxD is used to identifying the structural hole spanners in each granularity. Finally, using the measurement of the extent of node spanning structural holes to anal⁃ysis the effect of community structure under different granularity that influence the node’ s extent to span structural holes. Experimental results on public data and real data indicate that the extent of nodes to span structural holes namely the node’ s advantages will increase with the granularity get thinner.%目前研究者已提出一些基于社团结构的结构洞发现方法，然而不同粒度下社团划分结果使网络呈现层次化结构，影响社团结构中节点跨越结构洞的程度。本文基于网络社团划分思想提出一种分层网络的结构洞发现方法MG＿MaxD。首先，使用分层递阶社团划分算法（本文使用EAGLE算法），得到不同粒度的社团结构；然后，使用结构洞发现算法MG＿MaxD得到不同粒度下的结构洞占据者；最后，使用结构洞跨越程度指标分析不同粒度下的社团结构对节点跨越结构洞程度的影响。在公用和真实数据集上的实验结果表明节点跨越结构洞的程度即结构洞节点的优势将随着粒度的变细而增大。