0
$\begingroup$

I have a weighted undirected graph, where weight is the similarity and range from 0 to 1. I applied a density-based clustering method and get some clusters, with overlapping nodes (node can belong to more than one cluster), and with noise or outlier nodes ( node don't belong to any cluster).

Now, are there quality measures that works fine with this clustering result?

For example, one of measures I think about is Modularity Q, but I don't sure if it suitable for my graph or not. Also if other measures such as internal density, node betweeness, converge and entropy can be used here ?

Thank in advance.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

There was one recently proposed:

  • Moulavi, D., Jaskowiak, P. A., Campello, R. J. G. B., Zimek, A., & Sander, J. (2014).
    Density-based clustering validation.
    In Proceedings of the 14th SIAM International Conference on Data Mining (SDM), Philadelphia, PA.

Havn't read it in details yet, or tried it.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks a lot, it is really match my need except the overlapping feature, it is not supported. Do you have idea if it will affect the result if I use it or not ? And thanks again :) $\endgroup$
    – Ysak
    Commented Mar 12, 2016 at 6:18
  • 1
    $\begingroup$ I don't know. I believe none of the measures handles this case well. You may need to average per point over all its assignments. $\endgroup$
    – Has QUIT--Anony-Mousse
    Commented Mar 12, 2016 at 8:57

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.