IMPORTANCE NVARS

segaqaci — Tue, 21 Apr 2020 03:05:30 GMT

Hello.

I have a gradient boosting tree where I want to get a list of all the variables used in the model. I currently have "IMPORTANCE NVARS=40",but is there a way to set this to ALL or MAX, so that the same code will work on any dataset? Or do I need to sum up the number of variables I have in the data set and put that into a macro variable?

Here is the code:

PROC TREEBOOST data=&abt._
CATEGORICALBINS=&CATEGORICALBINS.
INTERVALBINS=&INTERVALBINS.
EXHAUSTIVE=&EXHAUSTIVE.
INTERVALDECIMALS=&INTERVALDECIMALS.
LEAFSIZE=&LEAFSIZE.
MAXBRANCHES=&MAXBRANCHES.
ITERATIONS=&ITERATIONS.
MINCATSIZE=&MINCATSIZE.
MISSING=&MISSING.
LEAFFRACTION=&LEAFFRACTION. 
SEED=12345
SHRINKAGE=&SHRINKAGE.
SPLITSIZE=&SPLITSIZE.
TRAINPROPORTION=&TRAINPROPORTION.
;
INPUT &charvars. &flgvars. /level= nominal ; 
INPUT &numvars. /level= interval ; 
TARGET target_flg /level=binary ;
IMPORTANCE NVARS=40 OUTFIT=imp_VARS out = imp20 ;
SUBSERIES BEST ;
CODE FILE="&save_path.\&save_file." ;
SAVE MODEL=GBoost_Test  FIT=fit IMPORTANCE=imp  RULES=rules; 
RUN;

Re: IMPORTANCE NVARS

WendyCzika — Tue, 21 Apr 2020 20:04:19 GMT

Doesn't look like there is an option like "ALL" for Nvars, but you could either set it to a very high value or like you said, use a macro variable to get the number of inputs, something like:

%let nclass = %sysfunc(countw(&charvars. &flgvars.));
%let ninterval =  %sysfunc(countw(&numvars.));
%let totnvars = %eval(&nclass + &ninterval);

proc treeboost...
importance nvars=&totnvars...;
...
run;

topic IMPORTANCE NVARS in SAS Data Science

IMPORTANCE NVARS

Re: IMPORTANCE NVARS