基于角色划分的动态社区挖掘算法研究

传统社区挖掘算法根据静态的网络拓扑结构进行分析,忽视了个体能动性对网络的影响。针对社会网络中的特殊节点进行研究,引入社区种子和联系者的概念,从个体主义和结构主义两个方面分析社会网络的形成与演化机制,提出了一种基于角色划分的动态社区挖掘算法。在人工网络和真实世界网络上进行了多次测试,并与G-N、快速G-N、Polish等算法进行了比较,结果表明,该算法明显优于G-N算法,且其挖掘到的社区都是强连通社区,具有较好的适应性和可扩展性。

Research of Dynamic Community Discovery Based on Role Assorted Thoughts

Traditional community discovery algorithms focus on the analysis of static topology structure of networks while ignoring the influence of individual activity on the formation of networks. This paper introduced the concept of community seed and liaison, and aiming at the special nodes, researched and analyzed the formation and evolution mecha- nism of social network from both individualism and structuralism perspectives, proposed a role assorted community dis- cowry algorithm. This paper tested the performance of this algorithm both on artificial network and real-world net- works and compared the results with GN, fast GN and Polish. Experimental results show that the results of role as- sorted algorithm are much better than GN algorithm, with great suitability and expandability. Besides, the discovery communities arc all strong connected communities.