Re: [igraph] How to see communities after community detection algorithm?

[Tamas Nepusz]

> My network has just 90 nodes, and hundreds of edges.
> I used the community detection provided by igraph, but the results are not
> good.
> Are they more fit for the large scale networks?>
The community detection methods are suitable for smaller networks as
well (e.g., the Zachary karate club network, a common testbed for
community detection algorithms contains only 34 nodes, and all the
algorithms in igraph work well on the karate club dataset). However,
note that if there is no definite community structure in your network
(which seems quite possible for me as your network is very dense), these
methods will nevertheless try to "find" something which may not be what
you are looking for. Are you sure that there should be some densely
connected regions in your network? A possible way to assess that is to
calculate the local transitivities (i.e. the local clustering
coefficients) for each vertex and check the distribution of the
transitivities; a clustered network should contain a few nodes
with transitivity values close to one, but not all of them should have a
transitivity close to zero (as that would indicate that your graph is
almost fully connected).

-- 
Tamas