PROC HPCLUS

[BenjaminDeKoven]

Is there a way to score new observations based on initial scoring using PROC HPCLUS?  I have seen posts on using FASTCLUS, but also interested in using HPCLUS.

Thank you,

Ben

[WendyCzika · Accepted solution]

PROC HPCLUS has a CODE statement to save the DATA step score code, so you can use that, then score your new data with it:

proc hpclus...;
 /* other statements */
  code file="hpclus.sas";
run;

data scored_data;
  set data_to_score;
  %inc "hpclus.sas";
run;

[BenjaminDeKoven]

Thank you for this excellent tip.

What are your suggestions to determine if a new observation's cluster scoring indicates that an additional cluster is now present compared to the training dataset?

[BenjaminDeKoven]

My dataset for training basically contains different platters / surfaces with defects at known locations (x and y).

How can I setup HPCLUS to use the distinct defects on multiple platters for traning?

Thank you,

Ben DeKoven