3
$\begingroup$

I have two ways to cluster nodes in a given graph, and now I want to compare my two clusters so to measure how "similar" the resulting clusters are. I have found some measures of similarity (Rand index, Variation of information, etc), but none of them takes into account the structure of the graph.

So my question is if exist some measure that considers distances in the graph?

Improve this question
$\endgroup$
1
  • $\begingroup$ Can you paste in a simple example dataset / pair of cluster outputs? $\endgroup$
    – gung - Reinstate Monica
    Commented Dec 14, 2017 at 17:57

1 Answer 1

Reset to default
1
$\begingroup$

Measures like Rand index are based on pairs of points.

A pair of points is essentially an edge.

These measures are based on the agreement on which object should be 'linked' and which should not be 'linked', so I would call of this family of measures "graph based". And for sure they will work on graph clusterings.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.