GRADBOOST Procedure

TRANSFERLEARN Statement

TRANSFERLEARN variable </ options>;

The TRANSFERLEARN statement enables you to train the gradient boosting model by using auxiliary data that are added to your training data. For more information, see the section Transfer Learning.

The TRANSFERLEARN statement names the variable whose value indicates whether the observation belongs to the training data or auxiliary data. A value between 0 and 1 indicates training data, and a value greater than or equal to 1 indicates auxiliary data. A value of 0 is given zero weight in the analysis. Observations that have missing values or negative values are ignored.

You can specify the following options:

BURN=number

specifies the number of trees to create before down-weighting any observation in the auxiliary data.

By default, BURN=0.

SHRINKAGE=number

specifies the number to apply as the weighting factor for down-weighting auxiliary data, where number must be between 0 and 1, exclusive.

By default, SHRINKAGE=0.9.

TRIMMING=number

specifies the number to use as a fraction of the distribution of gradients on the training data beyond which auxiliary observations are down-weighted, where number must be greater than 0 and less than or equal to 1/2.

By default, TRIMMING=0.01.

Last updated: August 06, 2026