How to use Normalized Mutual Information to evaluate overlapping community from igraph in python

Question

Given the increasing popularity of algorithms for overlapping clustering, in particular in social network analysis, quantitative measures are needed to measure the accuracy of a method. I need know how to use Normalized Mutual Information (NMI, available in igraph) to evaluate overlapping community from igraph in python.

This function: http://igraph.sourceforge.net/doc/python/

compare_communities(comm1, comm2, method='nmi', remove_none=False)

where

method = "nmi"

and comm1 and comm2 is a OverlappingClustering Object (http://cneurocvs.rmki.kfki.hu/igraph/doc/python/igraph.clustering.OverlappingClustering-class.html)

I tried this:

cl = igraph.Clustering([(0,), (0,), (0,1), (1,), (1,), ()])

But, it returns this error:

Traceback (most recent call last):
      File "multilevel_overlapping.py", line 94, in <module>
        cl = igraph.Clustering([(0,), (0,), (0,1), (1,), (1,), ()])
      File "/usr/lib/python2.7/dist-packages/igraph/clustering.py", line 92, in __init__
        self._len = max(m for m in self._membership if m is not None)+1
TypeError: can only concatenate tuple (not "int") to tuple

Also, tried this:

cl = igraph.OverlappingClustering([(0,), (0,), (0,1), (1,), (1,), ()])

But, it returns this error:

Traceback (most recent call last):
  File "multilevel_overlapping.py", line 94, in <module>
    cl = igraph.OverlappingClustering([(0,), (0,), (0,1), (1,), (1,), ()])
AttributeError: 'module' object has no attribute 'OverlappingClustering'

Would anyone can help me to use NMI function?

Answer 1

There seems to be a typo in the documentation. The class you're referring to is called OverlappingClustering not Clustering. Therefore try this:

 cl = OverlappingClustering([(0,), (0,), (0,1), (1,), (1,), ()])