Issue |
EPL
Volume 108, Number 6, December 2014
|
|
---|---|---|
Article Number | 68001 | |
Number of page(s) | 6 | |
Section | Interdisciplinary Physics and Related Areas of Science and Technology | |
DOI | https://doi.org/10.1209/0295-5075/108/68001 | |
Published online | 16 December 2014 |
Detecting overlapping communities in massive networks
CAS Key Laboratory of Network Data Science and Technology, Institute of Computing Technology, Chinese Academy of Sciences - Beijing 100190, China
(a) shenhuawei@ict.ac.cn (corresponding author)
Received: 23 September 2014
Accepted: 28 November 2014
Community detection is an essential work for network analysis. However, few methods could be used as off-the-shelf tools to detect communities in real-world networks for two main reasons: Real networks often contain millions of nodes or even hundreds of millions of nodes while most methods cannot handle networks at this scale. One node often belongs to multiple communities, posing another big challenge. In this paper, we circumvent the tricky problem of detecting overlapping communities using a two-stage framework, balancing efficiency and accuracy. Given a network, we first focus on efficiently finding its coarse-grained communities. Starting from them, we next obtain overlapping communities by optimizing a principled objective function. In this divide-and-conquer way, the framework achieves a much better performance than detecting overlapping communities from scratch. Extensive tests on synthetic and real networks demonstrate that it outperforms state-of-the-art methods in terms of both efficiency and accuracy.
PACS: 89.75.Fb – Structures and organization in complex systems / 89.75.Kd – Patterns / 05.65.+b – Self-organized systems
© EPLA, 2014
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.