PARTITION Subcommand (RBF command)

The PARTITION subcommand specifies the method of partitioning the active dataset into training, testing, and holdout samples. The training sample comprises the data records used to train the neural network. The testing sample is an independent set of data records used to track prediction error during training in order to prevent overtraining. The holdout sample is another independent set of data records used to assess the final neural network.

  • The partition can be defined by specifying the ratio of cases randomly assigned to each sample (training, testing, and holdout) or by a variable that assigns each case to the training, testing, or holdout sample.
  • If the PARTITION subcommand is not specified, then the default partition randomly assigns 70% of the cases to the training sample, 30% to the testing sample, and 0% to the holdout sample. If you want to specify a different random assignment, then you must specify new values for the TRAINING, TESTING, and HOLDOUT keywords. The value specified on each keyword gives the relative number of cases in the active dataset to assign to each sample. For example, /PARTITION TRAINING = 50 TESTING = 30 HOLDOUT = 20 is equivalent to /PARTITION TRAINING = 5 TESTING = 3 HOLDOUT = 2; both subcommands randomly assign 50% of the cases to the training sample, 30% to the testing sample, and 20% to the holdout sample.
  • If you want to be able to reproduce results based on the TRAINING, TESTING, and HOLDOUT keywords later, use the SET command to set the initialization value for the random number generator before running the RBF procedure.
  • Be aware of the relationship between rescaling and partitioning. See the topic RESCALE Subcommand (RBF command) for more information.
  • All partitioning is performed after listwise deletion of any cases with invalid data for any variable used by the procedure. See MISSING Subcommand (MLP command) for details about valid and invalid data.

TRAINING Keyword

The TRAINING keyword specifies the relative number of cases in the active dataset to randomly assign to the training sample. The value must be an integer greater than 0. The default (if the PARTITION subcommand is not specified) is 70.

TESTING Keyword

The TESTING keyword specifies the relative number of cases in the active dataset to randomly assign to the testing sample. The value must be an integer greater than 0. The default (if the PARTITION subcommand is not specified) is 30.

HOLDOUT Keyword

The HOLDOUT keyword specifies the relative number of cases in the active dataset to randomly assign to the holdout sample. The value must be an integer greater than 0. The default (if the PARTITION subcommand is not specified) is 0.

VARIABLE Keyword

The VARIABLE keyword specifies a variable that assigns each case in the active dataset to the training, testing, or holdout sample. Cases with a positive value on the variable are assigned to the training sample, cases with a value of 0, to the testing sample, and cases with a negative value, to the holdout sample. Cases with a system-missing value are excluded from the analysis. (Any user-missing values for the partition variable are always treated as valid.)

The variable may not be the dependent variable or any variable specified on the command line factor or covariate lists. The variable must be numeric.