中文摘要：

社团发现是复杂网络研究领域里一个极具挑战性的方向。特别地,对于现实世界中许多规模巨大的应用层拓扑,一些社团发现算法因为计算复杂度过高而不适用。另一些社团发现算法的实际性能还有待评估。为此,提出了可用于衡量社团发现算法实际应用价值的三个指标：（1）可扩展性,衡量算法能处理的应用层拓扑的规模;（2）准确度,衡量社团划分结果好坏;（3）敏感度,衡量社团划分结果的一致程度。在7个大规模的现实世界应用层拓扑图上,对三个针对大规模网络的社团发现算法（CNM算法、Wakita算法和Louvain算法）进行了比较分析。实验结果表明,Louvain算法在可扩展性上最优且所得划分结果准确度最高,CNM算法在敏感度上表现最好。

英文摘要：

Community detection is a very challenging research direction in the field of complex networks＇ research. In particular, as the large scale of real world＇s application layer topology, some community detection algorithms are not applicable because of high computational complexity. And some algorithms＇ actual performance remains to be evaluated. Therefore, the paper proposes three indicators that can be used to measure community detection algorithms＇ actual application value： （1） extensibility, used to measure the scale of application layer topology that can be dealt with; （2） accuracy, used to measure the partition result is good or bad ; （3） sensitivity, used to measure the partition result＇ s consistent. The paper uses seven large - scale application layer topologies to analyze three community detection algorithms for large - scale network, i.e. CNM algorithm, Wakita algorithm and Louvain algorithm. The results show that Louvain algorithm performs best on scalability and accuracy, and CNM algorithm＇s performance on sensitivity is best.