基于局部扩充的社区发现算法并行化研究

Research on an Improved Parallel Algorithm for Community Detection Based on Fitness Function

随着社交网络的快速发展,出现成员属于多个社区的现象。现有大多数算法研究重点为重叠社区(如LFM),其中对于高度重叠的社区发现研究依然是弱点。在LFM算法的基础上,提出极大团作为种子,自适应更新局部扩充质量优化函数参数α,并将扩充过程进行并行化的一种新型方法。经过理论证明和在人造数据图以及真实网络上试验,相比LFM,该算法在准确性和效率上均有较大提高。

With the rapid development of social network, the phenomenon which members belong to more than one community has coming. Most research of existing algorithms focus on overlapping communities（such as LFM）, however the research to highly overlapping community is still weak. On the basis of LFM algorithm, proposes a novel method which regards cliques as seed, adaptive updates parameters of local fitness function, and expanses seeds in parallel. When put to the task of identifying community in synthetic graphs and real social net-works, compared to the LFM, this algorithm is greatly improved in accuracy and efficiency.