The example implements a large scale overlapping community detection method based on local expansion via minimum one norm. The program adopts a local expansion method in order to identify the community members from a few exemplary seed members. The algorithm finds the community by seeking a sparse vector in the span of the local spectra such that the seeds are in its support. LEMON can achieve the highest detection accuracy among state-of-the-art proposals. The running time depends on the size of the community rather than that of the entire graph.
Here we use Amazon dataset (obtained from SNAP website) as an illustration. You may switch to other datasets with corresponding file format as well. Note that some parameters might need to be adjusted accordingly based on the properties of network under test.
@inproceedings{li2015uncovering,
title={Uncovering the small community structure in large networks: a local spectral approach},
author={Li, Yixuan and He, Kun and Bindel, David and Hopcroft, John E},
booktitle={Proceedings of the 24th international conference on world wide web},
pages={658--668},
year={2015},
organization={International World Wide Web Conferences Steering Committee}
}(may have to be independently installed)
936 communities with ground truth size >= 20.334863, edges: 9258724939$cd LEMON
$python LEMON.py -f ../example/amazon/graph -g ../example/amazon/community --sd ../example/amazon/seed --out output.txt-d: delimiter of input graph and community files [default: space]
-f: input network file [default: example/amazon/graph]
The format of a graph is edgelist, e.g::
1 2
1 3
1 4
...-g: input ground truth community file [default: example/amazon/community]
The format of a ground truth community is a space delimited line of node IDs , e.g:
1 4 8 14 20 21 22 # community 1
2 5 3 6 7 15 16 17 18 19 # community 2
9 10 11 12 13 23 # community 3--sd: initial seed set input file [default:example/amazon/seed]
The format of seed set is a single line of space delimited node IDs, e.g:
2 5--out: output file path [default: output.txt]
The output includes the detected community and the similarity between the detected community and ground truth community
(quantified by F1 score), e.g:
# detected community:
[2,5,3,6,7,15,16,17,18,19]
# F1 score: 1.0-c: minimum community size [default: 20]
-C: maximum community size [default: 100]
-e: expand step [default: 6]
The full list of command line options is available with $python LEMON.py --help