I've some data that I've clustered and calculate the similitude within my cluster centroids to spot which class centroids are close to which ones. My goal was to visualize this as a map of points. I thought that the best way was to represent the data as a complete graph where each node represent a centroid and the edges the similitude within the centroids. Plug this into a graph visualization software like gephi let the nodes pull an push each other, get a pretty picture and job done. But i couldn't make it work the graph collapsed all the time.

Maybe is just that my data is supposed to collapse, but here's how the similitude within my 36 nodes look like and to me, it seems that the graph visualization I've in mind should go somewhere. The other option is that i'm not familiar with the gephi software and i did not manage to make it work.

Do you have any suggestions? software? visualization technique? how would you show a more intuitive visualization of the similitude matrix?

thanksenter image description here


Have you tried using Multidimensional Scaling (MDS) to emebed the centroids in a 2 dimensional space for visualization?

This should visualize how the centroids related to each other. You could also then compute the Delauney Triangulation of the projected graph, to identify neighbors.

| cite | improve this answer | |
  • $\begingroup$ Actually you are right, instead of let the graph evolve etc. Just by projecting the first two first components of a PCA procedure would do the job. Thanks, I got obscured by my own thoughts and you pushed me out. $\endgroup$ – Sik Sep 3 '13 at 11:55

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.