The HPSVM Procedure

SELECT Statement

SELECT <select-options>;

You can use the SELECT statement in combination with the PENALTY statement to generate the best penalty value. The best penalty is selected through the cross validation statistics.

The SELECT statement specifies how to subdivide the training data into k parts when you request k-fold cross validation, where k is an integer between 2 and 20 inclusive.

You can specify the following select-options:

CV=BLOCK | SPLIT | RANDOM | VALIDATESET

specifies the partition method. You can specify the following values:

BLOCK: requests that parts be formed of k equal-sized blocks of consecutive training observations.
SPLIT: requests that the ith part consist of training observations i, i+k, i+2k, and so on.
RANDOM: assigns each training observation randomly to one of the k parts.
VALIDATESET: requests that the penalty be selected by using the validation set that is specified in the PARTITION statement.

For k-fold cross validation, the default value is CV=RANDOM.

FOLD=k

specifies the number of folds for k-fold cross validation. If you specify CV=VALIDATESET, then the FOLD= suboption is not needed. In this case, the penalty is selected by using the validation set that is specified in the PARTITION statement.

SEED=number

specifies the seed for the random number generator if CV=RANDOM. The value of number must be a nonnegative integer. By default, SEED=0.

The penalty selection process is time-consuming. For k-fold cross validation, the total number of training is k times the number of the penalties. For example, if FOLD=10 and the number of penalties is 10, then the total number of training is equal to 100.

The SELECT statement is optional.

Last updated: May 25, 2022