SUPERLEARNER Procedure

PROC SUPERLEARNER Statement

PROC SUPERLEARNER <options>;

The PROC SUPERLEARNER statement invokes the procedure. Table 1 summarizes the available options in this statement.

Table 1: PROC SUPERLEARNER Statement Options

Option	Description
APPLYROWORDER	Uses group and order information from the input data table
DATA=	Specifies the input data table
METHOD=	Specifies the method to use for fitting the meta-learner model
RESTORE=	Specifies the input item store
SEED=	Specifies the seed for the pseudorandom number generator

You can specify the following options:

APPLYROWORDER

uses group and order information from the input data table. You can add this information to the input table by using the partition action in the table action set. For more information, see the section The APPLYROWORDER Option in Chapter 2, Shared Concepts. This option is applicable to models that require the simulation of latent variables; by default, it is not enabled. If you specify this option but the table does not contain group and order information, the procedure terminates with an error.

DATA=libref.data-table

names the input data table for PROC SUPERLEARNER to use. The default is the most recently created data table. libref.data-table is a two-level name, where

libref: refers to a collection of information that is defined in the LIBNAME statement and includes the library, which includes a path to the data, and a session identifier, which defaults to the active session but which can be explicitly defined in the LIBNAME statement. For more information about libref, see the section Using CAS Sessions and CAS Engine Librefs.
data-table: specifies the name of the input data table.

METHOD=method<(method-option)>

specifies the meta-learning method to be used to estimate the coefficients of the super learner model. By specifying a meta-learning method, you specify an algorithm that is used to combine the individual base learner models in the library. For more information, see the section Meta-learning Methods. You can specify one of the following methods:

CCAUC: estimates the coefficients by maximizing the convex-constrained area under the ROC curve (AUC). This option is available only if the response variable is binary.
CCLOGLIK: estimates the coefficients by maximizing the convex-constrained binomial log likelihood. This option is available only if the response variable is binary.
CCLS: estimates the coefficients by fitting a convex-constrained least squares regression with no intercept.
CVSELECTOR: estimates the coefficients by assigning a coefficient of 1 to a single base learner model on the basis of cross-validated performance and setting coefficients of all other base learner models to 0.

By default, METHOD=CCLS if the response variable is continuous and METHOD=CCLOGLIK if the response variable is binary.

You can specify the following method-option when METHOD=CVSELECTOR:

LOSS=keyword

specifies the loss function to use for computing cross-validated risks. You can specify the following keywords:

AUC: uses the area under the ROC curve loss function. This option is available only if the response variable is binary.
LOG: uses the log loss function. This option is available only if the response variable is binary.
QUADRATIC: uses the quadratic loss function.

By default, LOSS=QUADRATIC if the response variable is continuous and LOSS=LOG if the response variable is binary.

RESTORE=libref.data-table

specifies the name of the item store that contains a model that is fitted and stored from a previous analysis. The item store is created by a STORE statement from a previous PROC SUPERLEARNER call or by a store parameter that is specified in a previous slTrain action call. libref.data-table is a two-level name, where libref refers to the library, and data-table specifies the name of the input data table. For more information about this two-level name, see the section Using CAS Sessions and CAS Engine Librefs.

You can use the previously fitted model to either score data or compute predictions at fixed values of the predictors. For each task, you use the DATA= option to specify the input data table to be used. To score the input data, you specify the OUTPUT statement to create a data table that contains the observationwise predictions. To compute predictions at fixed values of the predictors, you specify one or more MARGIN statements and specify the MARGINPRED option in the OUTPUT statement. You can also request observationwise predictions from each base learner model that you use to train the super learner model. To do so, specify the LEARNERPRED option in the OUTPUT statement.

Because the previously fitted model is stored and retrieved for the analysis, you cannot include statements and options that are specific to training a model. In particular, the BASELEARNER, CROSSVALIDATION, INPUT, STORE, and TARGET statements are not available when you specify the RESTORE= option.

SEED=n

specifies the initial seed to start the pseudorandom number generator for k-fold partitioning and base learner model building, where n is an integer. If you omit the SEED= option or if n is negative or 0, the procedure uses the time of day from the computer’s clock to obtain the initial seed.

Last updated: June 22, 2026