topic Re: PCA using K-mean in Statistical Procedures

PCA using K-mean

Asudipta — Sun, 18 Jan 2015 02:39:28 GMT

Hello everyone, Can you please suggest how to use PCA using K means in SAS? Are there separate procedures for the classifier, evaluation, and optimization of this algorithm?

Re: PCA using K-mean

PGStats — Sun, 18 Jan 2015 03:28:59 GMT

You seem to be asking about discriminant analysis, not principal component analysis... For k-nearest-neighbors based discrimination, look at proc discrim with option method=npar. For variable selection, there is proc stepdisc but it doesn't offer non-parametric discrimination.

Re: PCA using K-mean

Asudipta — Sun, 18 Jan 2015 07:02:55 GMT

I meant PCA using K-mean as both the algorithm are based on similar principals. I found it here
http://http://epubs.siam.org/doi/pdf/10.1137/1.9781611972740.54.

Thank you for sharing.

Re: PCA using K-mean

PGStats — Sun, 18 Jan 2015 22:25:35 GMT

Which of the three experiments described in that paper do you wish to implement? - PG

Re: PCA using K-mean

Asudipta — Mon, 19 Jan 2015 01:54:45 GMT

How is that a concern? I was trying to reiterate the fact that it was PCA and not DA. Can we come back to the original question now?Do you have an example to implement PCA using Kmeans? Thank you

Re: PCA using K-mean

Reeza — Mon, 19 Jan 2015 04:44:58 GMT

I think you need a combination of proc cluster and proc princomp.