Betweenness Centrality calculation?

[iiyonite]

Hi,

can anyone explain to me how gephi is calculating the betweenness centrality figures,
because I am unable to explain the oddity in my graphs.

In my particular case all edges are weighted with 1, as I want to highlight the number of edges without the modulary class (blocs) as a measure of centrality in the whole system.

The case I am unable to explain is the huge difference between Cameroon and Israel in my graph.
Cameroon is ranked at a betweenness value of 794.99 and a weighted degree of 32.0
Israel on the other hand has a betweenness value of only 125.01 with a weighted degree of 40.0

attached are two pictures which visualize this oddity ... to me it seems that Israel is more central than Cameroon ... but gephi thinks otherwise.


any ideas on why this huge betweenness centrality value difference? and how to rectify it?
thanks in advance

[pegerp]

Hi iiyonite.

I don't know the details of the algorithm implementation but my intuition is that Gephi is using the random walk closeness centrality, check the description at http://en.wikipedia.org/wiki/Random_wal ... centrality

So intuitively in your data it is more likely that a random walk goes through Cameroon rather than Israel even though Israel is more connected to its neighbours. So when neighbours' neighbours and etc. are taken into account Cameroon is more central than Israel.

To size nodes according to their degree you can compute the "Average degree" in the Statistics panel. Then use Degree column as the size parameter for nodes. If this is what you actually want you don't need to compute the centrality value at all.

What is the data representing? I observe that many Western countries are quite small nodes there. Developing countries are larger in size. In this respect it's quite intuitive that Israel would be a small node too.

[jlarrosa]

Hi iiyonite.

I don't know the details of the algorithm implementation but my intuition is that Gephi is using the random walk closeness centrality, check the description at http://en.wikipedia.org/wiki/Random_wal ... centrality

Hi, Pegerp, or anybody in the Gephi developing team, could anybody confirm if Gephi's betweeness calculation is using random walk closeness centrality algorithm?

Greetings, Juan

[ardalan3]

Hi, as reported below the result graphs the utilized algorithm is : Ulrik Brandes, A Faster Algorithm for Betweenness Centrality, in Journal of Mathematical Sociology 25(2):163-177, (2001)