NNET Procedure
Syntax: NNET Procedure
The following statements are available in the NNET procedure:
PROC NNET <options>;
ARCHITECTURE architecture-options;
AUTOTUNE <options>;
CODE <options>;
CROSSVALIDATION <options>;
FREQ variable;
HIDDEN number </options>;
INPUT variables </ LEVEL=INTERVAL NOMINAL>;
MITIGATEBIAS mitigate-options;
OPTIMIZATION <options>;
PARTITION <partition-options>;
SAVESTATE RSTORE=libref.data-table;
OUTPUT OUT=libref.data-table <option>;
TARGET variables </options>;
TRAIN OUTMODEL=libref.data-table <options>;
WEIGHT variable;
When you train a neural network, the PROC NNET, INPUT, TARGET, and TRAIN statements are required. At least one HIDDEN statement is required unless you use the GLIM architecture; in that case the HIDDEN statement is not allowed.
When you use a previously trained neural network to score a data table, only the PROC NNET, OUTPUT, and CODE statements are allowed.