Running Modularity statistics on the same data set twice generates different numbers of communities?

[yuling]

Hi All,

I was hoping that the communities generated from Modularity statistics could be same of similar if I click the button twice. However, with the same selections of "Randomize", "Use weights", and "Resolution", it first generated 56 communities, and then 79 communities...any idea?

Thanks,
YuLing

[eduramiba]

Hi,

Modularity is non-deterministic, it depends on the randomization, which chooses what's the starting node. Try to disable it.

[Max Viana]

I try to do that (disable random mode), but it's still generating different results.

Any idea?

[eduramiba]

Hi,
Randomization makes the algorithm start in a random node or not. So, several runs in the same workspace should work the same.
But it's possible that due to the graph structure (graphstore) the first node in the graph might change if you load the same file several times.