0
$\begingroup$

I am evaluating a few non-supervised clustering algorithms. One of the questions that I was asked was the internal quality for each algorithms. Any suggestions?

Secondly, what would be an appropriate metric for calculating similarity across the multiple clusters? Most common metrics such as Rand Index only compares between two clusters. Thanks!

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

There are probably around 30-40 such consistency measures.

The same holds for comparing clusters. ARI is one of the most popular.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks for your comments @Anony-Mousse. I've tried avg silhouette but it wasn't 'sensitive' (for a lack of a better word) enough to pick out the outliers. Perhaps I just have to keep trying different measures... ARI is great for pair-wise comparison, but I was wondering is there are ways to compare multiple groups? $\endgroup$
    – SplitInf
    Commented Jun 13, 2017 at 14:52
  • $\begingroup$ I don't see how you would use silhouette? And from a pairwise comparison you can trivially go to an n-way distance matrix? $\endgroup$
    – Has QUIT--Anony-Mousse
    Commented Jun 13, 2017 at 15:14

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.