CATTRANSFORM Procedure

The CATTRANSFORM procedure transforms categorical variables in SAS Viya Workbench. It supports binning and encoding methods that are commonly performed as part of the data preprocessing step in creating machine learning models.

Categorical variables that have high cardinality can slow the computing performance of machine learning algorithms and increase the risk of model overfitting. PROC CATTRANSFORM supports several binning methods, both supervised and unsupervised, that enable you to reduce cardinality. These methods include grouping rare levels, classification tree binning, regression tree binning, and weight-of-evidence binning. The procedure also supports one-hot encoding by transforming the values of categorical variables to numeric values that start with the number 1.

PROC CATTRANSFORM facilitates transformation scoring of new data by using either an analytic store or scoring code in a SAS DATA step.