点击次数：

论文类型： 会议论文

发表时间： 2011-04-22

收录刊物： EI、CPCI-S、Scopus

卷号： 6637

页面范围： 287-298

关键字： Community structure; Local modularity; Local core

摘要： To identify global community structure in networks is a great challenge that requires complete information of graphs, which is not feasible for some large networks, e.g. the World Wide Web. Recently, local algorithms have been proposed to extract communities in nearly linear time, which just require a small part of the graphs. However, their results, largely depending on the starting vertex, are not stable. In this paper, we propose a local modularity method for extracting local communities from local cores instead of random vertices. This approach firstly extracts a large enough local core with a heuristic strategy. Then, it detects the corresponding local community by optimizing local modularity, and finally removes outliers based on introversion. Experiment results indicate that, compared with previous algorithms, our method can extract stable meaningful communities with higher quality.