topic How could I validate clusters from k-means in SAS Data Science

How could I validate clusters from k-means

fri0 — Fri, 14 Aug 2015 21:02:56 GMT

Hi, I have a problem with clustering. I don't know how to validate that clusters are not similar. Clusters look like they are overlayed in the proximity graph. I'm attaching the file.

Also, anyone could explain me more abaout this graph, for example, what is the dimension meaning?

Thanks in advance and sorry for my english.

PD. I couldn't upload the file, but it's a proximity graph where 3 clusters has the centroid in the same place practically.

Re: How could I validate clusters from k-means

Ksharp — Sat, 15 Aug 2015 11:33:59 GMT

I can think two way:

1) Cross-validate test

2) Use ANOVA test - Check PROC GLM or PROC ANOVA if your design is completed balanced design .

Re: How could I validate clusters from k-means

M_Maldonado — Sun, 16 Aug 2015 23:45:20 GMT

fri0,

One more thing to do is to connect a Segment Profiler node after your cluster node. Use the plots in the results to visually confirm that each cluster has a different distribution for certain variables. Press F1 to open the SAS Enterprise Miner Reference Help and take a look at the section for the segment profiler node.

I hope it helps,

Miguel

Re: How could I validate clusters from k-means

fri0 — Mon, 17 Aug 2015 18:53:20 GMT

Hi, thank your very much for your answers.

I have another problem. I have been questioned about the cluster proximity plot. I have only 6 cluster, but three of them are very close, almost overlapped. Further more, you could see an asterisk with a circle for each cluster, but in my graph, you could only see the asterisk. I really need help ton understand that graphic and could interpret it.

Thanks in advance and sorry for my english.