***** To join INSNA, visit http://www.insna.org ***** What you have explained here is the Jaccard index (http://en.wikipedia.org/wiki/Jaccard_index), which is very closely related to the cosine similarity measure (http://en.wikipedia.org/wiki/Cosine_similarity). Both measures are used for measuring how similar two samples (sets, clusters, etc) are.

Best
Aziz

Abedelaziz Mohaisen
Ph.D. Student, UMN
http://www.cs.umn.edu/~mohaisen

On Mon, Feb 21, 2011 at 9:07 PM, John McCreery wrote:
***** To join INSNA, visit http://www.insna.org ***** A few months ago I stumbled across a simple measure of cluster overlap. Assume two clusters, A and B, that contain a total of n nodes of which m nodes belong to both clusters or, in other words, n is the number of nodes in the union of A and B and m the number of nodes in the intersection of A and B. The measure is something along the lines of  overlap=m/n=nodes in intersection/nodes in union. My problem is (1) that I have forgotten where I found this measure and (2) am not 100% sure that it was as simple as I recall. If you recognize this problem, can point me to a source or provide me with a correct or better formula, your help will be much appreciated.

John

--
John McCreery
The Word Works, Ltd., Yokohama, JAPAN
Tel. +81-45-314-9324