0
$\begingroup$

I've recently run a very large data set through a multidimensional scaling analysis and am attempting to cluster the results into groups. I've read a few papers that utilize hierarchical clustering to accomplish this, however my data set is still rather large which is making Hclustering unfeasible. I have instead decided to use k-means, but I have yet to find any articles that either support this method or have used it.

Is k-means clustering a feasible option for my objective?

| cite | improve this question | |
$\endgroup$
1
$\begingroup$

Yes and no. There are some procedures that work in this way: you fix k=100. You run such a k-means, then you take the groups and mount a hierarchy. I found this method implemented in an old software called SPAD.

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Could you elaborate a little more on that process? To my understanding what you are saying is that I should run a k-means with a fixed k, then run those clusters through a hierarchical model? $\endgroup$ – r_user Feb 25 '19 at 12:43
  • 1
    $\begingroup$ It is a hybrid between k-means and hierarchical clustering. Once you obtain a k(let's say k=100), you have 100 clusters. With these 100 groups, you decide what linkage (complete, average, minimum, Ward, etc.) you want and mount a hierchy starting from the 100 clusters. $\endgroup$ – antonio irpino Mar 22 '19 at 10:15

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.