Finding Statistically Significant Communities in Networks
The communities are those of an LFR benchmark graph (undirected, unweighted and without overlapping clusters), with , , , . The cluster size ranges from to vertices. The noise comes by adding vertices which are randomly linked to the existing vertices, via preferential attachment. The test consists in checking whether the community finding algorithm at study (here OSLOM, Infomap and COPRA) is able to find the communities of the planted partition of the LFR benchmark and to recognize as homeless the other vertices.