Build and Test Classification Model Procedure

This procedure is generated from an activity that successfully builds and tests a classification model. The procedure builds the model and optionally tests the built model. The number of input parameters for additional tables (and schemas) can vary depending on the number of additional tables passed to the build activity. All parameters are initialized to the default values that were used by the build activity. If you use input parameters that are not the default parameters, you must ensure the new data is compatible with the original data used to build the model.

No Test

You can generate code for a classification activity where the Test step was not run. In this case, the default parameters for the test results (confusion matrix name, lift result name, ROC result name, and test metric name) are all NULL. If you want to test the model, you must specify the test result names.

Syntax

PROCEDURE build_class_procedure_name(
               case_table            IN VARCHAR2 DEFAULT activity_default,
               table_1               IN VARCHAR2 DEFAULT activity_default,
               table_n               IN VARCHAR2 DEFAULT activity_default,
               model_name            IN VARCHAR2 DEFAULT activity_default,
               confusion_matrix_name IN VARCHAR2 DEFAULT activity_default,
               lift_result_name      IN VARCHAR2 DEFAULT activity_default,
               roc_result_name       IN VARCHAR2 DEFAULT activity_default,
               test_metric_name      IN VARCHAR2 DEFAULT activity_default,
               feature_table         IN VARCHAR2 DEFAULT activity_default,
               mapping_table         IN VARCHAR2 DEFAULT activity_default,
               drop_output           IN BOOLEAN DEFAULT user_defined);

Parameters and Defaults

The procedure has two kinds of defaults:

Table 1 Build and Test Classification Model Procedure Parameters

Parameter Description
build_class_procedure_name The name specified for the procedure when the code was generated.
table_1 The name of the first additional table. The name is in the form schema_name.table_name. This parameter is optional.
table_n The name of the n-th additional table. The name is in the form schema_name.table_name. This parameter is optional.
model_name The name of the model created by the procedure; 25 or fewer characters in length.
confusion_matrix_name The name of the confusion matrix created by the procedure. NULL if the Test step in the activity was not run.
lift_result_name The name of the output lift result created by the procedure. NULL if the Test step in the activity was not run.
roc_result_name The name of the output Receiver Operating Characteristics (ROC) result created by the procedure. NULL if the Test step in the activity was not run.
test_metric_name The name of the output test metrics created by the procedure. NULL if the Test step in the activity was not run.
mapping_table Feature frequencies table generated by the text transform. Utilized by the apply and test procedures to regenerate a similar text transformation.
feature_table Feature ID to text mapping table generated by the text transform. Can be used to translate model detail output containing columns affected by a text transformation.
drop_output A flag indicating whether to drop the model and the test results if they already exist. The value of this flag was specified when the code was generated.

Usage Notes

This procedure includes all necessary data preparation, data transformations, model build settings, and model test settings. The procedure builds and tests a classification model.

If the test step is not completed in the activity, the procedure will contain the test code but the test code will be skipped when the procedure executes. If the test step is skipped, all the test result names default to NULL.

To generate any test result, you must specify the test_metric_name. The procedure creates all the results that are named. If you specify test_metric_name, the procedure calculates the predictive confidence value and appends the value to the output test metric table. If a confusion matrix is generated, the procedure also calculates the average accuracy and appends the value to the output test metric table.

If you don't specify a target value in the build activity, the lift and ROC are not generated even if you specify names for them.

If the model and test results already exist and drop_output is set to FALSE, the procedure will raise an exception to the caller and exit.

The feature table generated by the Build procedure is used in the Apply procedure.

You can supply a feature_table name different from the default name, but you must supply the same name in both Build and Apply procedures.