The HP4SCORE Procedure

IMPORTANCE Statement

IMPORTANCE FILE=model-file-name VARS=(variables) <importance-options> ;

The IMPORTANCE statement computes variable importance based on the random branch assignments (RBA) method. The output data set contains modified predictions of the input observations from which the importance of an individual observation can be calculated. The average importance is output in the variable importance table. For more information about RBA, see the section Random Branch Assignments Variable Importance.

You must specify the following arguments:

FILE=model-file-name: specifies either the file reference or the full path and member name of the valid model file that was created by PROC HPFOREST.
VARS=(variables): species one or more variables for which to calculate the variable importance. Dashes and colons that represent ranges of variables are not recognized in this context.

You can also specify the following importance-options:

MAXDEPTH=<n>: produces predictions from trees that are pruned to a depth of n. (The trees are not pruned by default.)
NTREES=<n>: produces predictions from the first n trees only. If this option is omitted, then PROC HP4SCORE uses all the trees in the model for predictions. Scoring with fewer trees can sometimes increase the speed without significantly reducing the accuracy.
OUT=<libref.>SAS-data-set: names the output SAS data set to contain the scored results. If this option is omitted, then no output data set is created.

Last updated: May 25, 2022