Wakita Tsurumi for directed graphs

Sep 8, 2012 at 9:59 AM

Dear All,

Mine is probably more a methodlogical question, so apologies in advance if that's not the best place to ask it. However I could not find a clear answer to that anywhere else.

I am comparing two networks (not necessarily involving the same actors). One relationship is directed, the other is not. In both networks I want to detect subgroups and I am using the Wakita Tsurumi algorithm. 

Can I use this on both directed and undirected graphs with no worries? Reading some papers I have understood that directionality in networks should call for adaptation of the algorithm, so I wonder if there is anything I should be cautious about or anything I should do before starting the analysis. Or anything which will impact the interpretation when comparing the two networks associated with the directionality issue?

Thanks in advance for your attention

Riccardo

Sep 10, 2012 at 5:56 PM

Riccardo:

NodeXL doesn't pay attention to the graph's directedness when you group by cluster using the Wakita-Tsurumi algorithm.  The Wakita-Tsurumi paper that describes the algorithm, "Finding Community Structure in Mega-scale Social Networks," doesn't mention directedness, so it is not a factor when clusters are determined.  You will get exactly the same results whether the graph is directed or undirected.

-- Tony