Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

Score a dataset with SVM

New Contributor
Posts: 4

Score a dataset with SVM

I have trained an SVM model with great success. Now I bring in another dataset for scoring. It has all the attributes except the target variable.

However, I got an error below. Does it mean that SVM scoring function is not provided in SAS EM 12.1? Thank you.

19570 +proc svmscore data= &em_score_output inest=EMWS1.SVM_OUTEST infit=EMWS1.SVM_SVMFITSTAT inclass=EMWS1.SVM_OUTCLASS

19571 +              out = &em_score_output(rename=(_F_=F_target_ind _I_=I_target_ind))

19572 +;

19573 +run;

ERROR: Variable Replicate not found on data set.

WARNING: This is an experimental release of the SVMSCORE procedure. It has not had the same level of testing as other components of the SAS system.

NOTE: The SAS System stopped processing this step because of errors. WARNING: The data set EMWS1.SVM_SCORE may be incomplete.  When this step was stopped there were 0 observations and 0 variables. WARNING: Data set EMWS1.SVM_SCORE was not replaced because this step was stopped.

Trusted Advisor
Posts: 1,204

Re: Score a dataset with SVM

Never used SVM but from the above detail seems like your training algorithm requires "Replicate" variable which is missing in scoring dataset.

SAS Employee
Posts: 15

Re: Score a dataset with SVM

What stat@sas says is correct. Without seeing more, the error indicates that you used the variable "Replicate" on the training data set. If this variable is not in the data set you wish to score, you receive that error that you have shown.

One way to avoid this is to retrain your model, but remove the variable "Replicate" from the training data set (or just don't tell svm to use that variable). Then when you go to score, the "proc svmscore" will not look for that variable on the scoring data set.

This is about as much advice I can offer without knowing more about your situation. I hope it helps.

Ask a Question
Discussion stats
  • 2 replies
  • 3 in conversation