SVMACHINE Procedure

In this example, support vector regression training is performed. For the sake of simplicity, the same data set, mylib.dmagecr, is used. Instead of using the binary target good_bad, amount is chosen as the interval target. The insensitive loss parameter epsilon is set to 5.0, and the LEVEL=INTERVAL option in the TARGET statement requests support vector regression training.

The following statements execute SVR algorithm on the mylib.dmagecr data table and produce Output 39.3.1 and Output 39.3.2:

proc svmachine data=mylib.dmagecr epsilon=5.0 noscale;
   input checking history purpose savings employed marital coapp
         property other job housing telephon foreign/level=nominal;
   input duration installp resident existcr depends age/level=interval;
   target amount / level=INTERVAL;
run;

The "Model Information" table is displayed in Output 39.3.1.

The "Training Results" table is displayed in Output 39.3.2.