Detecting highly overlapping communities with Model-based Overlapping Seed ExpanSion (MOSES)
MOSES is a scalable, overlapping, community finding algorithm. In empirical social networks, people are often members of many communities. In Facebook datasets we have seen that the average person is in seven communities, a quantity already observed in the literature on ego-networks. MOSES can find these heavily overlapping communities for every person in networks with millions of links within hours, an advance on previous methods which applied only to ego-networks or to large, weakly overlapping, datasets.
A script is currently under review for ASONAM 2010: Aaron McDaid, Neil Hurley. Detecting highly overlapping communities with Model-based Overlapping Seed Expansion
This software is made available under the Apache License version 2.0. The software is provided "as is" without express or implied warranty. If you use this software in your research, we encourage you to cite the associated paper:
- Aaron McDaid, Neil Hurley. Detecting highly overlapping communities with Model-based Overlapping Seed Expansion. ASONAM 2010.
The Facebook datasets analyzed in our paper. That dataset made available by Traud et al in Community Structure in Online Collegiate Social Networks. We also made extensive use of synthetic benchmarks (Fortunato and Lancichinetti. Benchmarks for testing community detection algorithms on directed and weighted graphs with overlapping communities).