I applied number of methods of clustering, and I want to evaluate these different methods using Dunn index, in this method I have to calculate the distance among clusters and among points in clusters.

My question is: if my algorithm is clustering users depending on their sequences, i.e., each user has sequences, then the algorithm measures the similarity among sequences using seq. alignment technique, then cluster them. Sequence alignment tool considers two users most similar when they have max score. If my clusters are id's of users or sequences of users, how do I compute the distance among clusters and among sequences within a cluster?

| cite | improve this question | | | | |

You supposedly should be able to use your sequence alignment measure as distance function for the Dunn index.

IIRC the between-cluster distances are derived from the distances of individual cluster members.

| cite | improve this answer | | | | |
  • $\begingroup$ I agree with @Anony-Mousse. Dunn index requires only distance matrix as input. But note that there are also some cluster validity indices that require original data (for example those that compute cluster centers). $\endgroup$ – Miroslav Sabo Oct 12 '12 at 7:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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