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

Error proc treeboost measure=LIFT

Reply
Highlighted
New Contributor
Posts: 2

Error proc treeboost measure=LIFT

Bonjour à tous,

 

J'ai un message d'erreur étrange lorsque je souhaite lancer ma proc treeboost en précisant measure=LIFT

 

Message d'erreur : ERROR: Internal Error: ABNORMAL return ignored.

 

Voici mon code :

 

PROC TREEBOOST DATA=sasuser.train3 
LEAFSIZE=100
MAXBRANCH=3
ITERATIONS=50 
MINCATSIZE=15
SEED=05042017
SHRINKAGE=0.1
SPLITSIZE=100
TRAINPROPORTION=0.5
;
INPUT &VAR_FAI. / LEVEL=INTERVAL INTERVALDECIMALS=5 ORDER=DESCENDING;
INPUT &VARTEXT. / LEVEL=NOMINAL;
TARGET FLG_CHURN /LEVEL=BINARY;
IMPORTANCE OUTFIT=sasuser.BASE_VARS50;
SUBSERIES BEST;
ASSESS measure=LIFT proportion=0.20;
SAVE IMPORTANCE=sasuser.gbm01varimp50 MODEL=sasuser.gbm01modl50;
RUN;

Merci d'avance pour votre aide

 

SAS Employee
Posts: 229

Re: Error proc treeboost measure=LIFT

The TREEBOOST procedure is a non-HP procedure used by the Gradient Boosting node in SAS Enterprise Miner and is available only to those who have SAS Enterprise Miner licensed.  Unfortunately, there is no support available for the direct use of these non-HP procedures since they were not intended to be used directly.   If you have SAS Enterprise Miner licensed, you can run the Gradient Boosting node to see how TREEBOOST is used. 

 

Having said that, the Gradient Boosting node examples I ran always showed

 

         proc arbor proc=treeboost data=dataset ....

 

so it might just be the way you are trying to use TREEBOOST.   

 

Hope this helps!

Doug

 

 

 

New Contributor
Posts: 2

Re: Error proc treeboost measure=LIFT

Posted in reply to DougWielenga

Hi Doug,

 

Thank you for your answer.

I forgave to precise this procedure it's totally OK if I don't precise "measure=LIFT" and the proportion.

That's why I don't understand why the measure option doen't run...

 

 

Ask a Question
Discussion stats
  • 2 replies
  • 237 views
  • 0 likes
  • 2 in conversation