This question already has an answer here:

I learned that it's common to do dimensionality reduction before clustering. But,is there any situation that it is better to do clustering first then do dimensionality reduction?

Thank you so much ^^


marked as duplicate by Anony-Mousse clustering Jan 14 '17 at 13:36

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • 1
    $\begingroup$ We do not always do or need dimensionality reduction prior clustering. Reducing dimensions helps against curse-of-dimensionality problem of which euclidean distance, for example, suffers. On the other hand, important cluster separation might sometimes take place in dimensions with weak variance, so things like PCA may be somewhat dangerous to do. It is double-edged problem. $\endgroup$ – ttnphns Jan 14 '17 at 9:14
  • $\begingroup$ Thank you so much. I think I chose the wrong direction. : ) $\endgroup$ – user145177 Jan 15 '17 at 7:54

Clustering generally depends on some sort of distance measure. Points near each other are in the same cluster; points far apart are in different clusters. But in high dimensional spaces, distance measures do not work very well. There is a long and excellent discussion of that Here. You reduce the number of dimensions first so that your distance metric will make sense.

  • $\begingroup$ And...I know clustering method like K-means is not good in high dimensional spaces because of computation complexity, but I still wonder is there any special cases, it is better to do clustering before dimension reduction? Is it possible? Thank you.^^ $\endgroup$ – user145177 Jan 14 '17 at 5:04

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