复杂合著网络中的重叠社团发现与可视化

Overlapping Community Discovery and Visualization in Complex Co-authorship Network

指出随着互联网的发展和社交网络的广泛应用,学者之间的合作途径越来越多,学者具备多重的社团属性;但是,传统的基于聚类或模块度优化等社团划分方法往往将学者划分到唯一的社团。基于Salton方法构建合著网络,使用加权的链接聚类算法实现社团的聚类划分,该方法将节点间的边作为聚类对象,采用凝聚式层次聚类进行社团划分。因节点属于不同的边,因而可以归属于不同的社团,最终得到的社团可以部分重叠。为检验方法的有效性,使用基于C-DPLP的合著网络构建系统获取数据,构建合著网络;然后使用加权的链接聚类进行社团发现和可视化。结果表明,该方法能有效地发现部分重叠的合著社团,且社团的意义比较明确。

With the development of Internet and SNS, there are so many ways to can＇y out academic cooperation between scholars that one scholar may belong to several communities. However, every scholar must belong to only one community when using traditional community disc：overy methods. In this paper, co-authorship network is constructed based Salton relation strength and community discovery in co-authorship network is performed based on weighted link clustering. In this method, weighted link is regarded as clustering object and one node may belong to several elusters. As a result, overlapping co-authorship community can be identified. To validate this method, several co-authorship networks were constructed based on Co-authorship Network Construct System developed by us and then experiments were earrifled out on co-authorship networks using weighted link clustering. The result shows that our method can find explicit co-authorship communities with partial overlapping and each eommunity is more characteristic compared with others.