If you want to monitor
the performance of a model for which you no longer have the score
code, you can import a model without SAS score code into the score.sas
file. If the performance data set contains the predicted values, the
score.sas file can be empty.
To monitor the performance
of a model without score code, follow these steps:
-
Prepare the following
model files:
-
XML file that defines the model
input variables (inputvar.xml)
-
XML file that defines the model
output variables (outputvar.xml)
-
XML file that defines the model
target variables (targetvar.xml)
-
empty SAS score code file (score.sas)
-
Create a project that
has a model function type of
Classification or
Prediction,
and create a version. You can skip this step if you have already created
a project and version.
-
If it is a project that
has a model function property value of
Classification,
verify that the following project properties are set:
-
Training Target Variable (for example,
bad)
-
Target Event Value (for example,
1)
-
Class Target Level as
Binary
(SAS
Model Manager 12.1 or later)
-
Output Event Probability Variable
(for example,
score
)
If it is a project that
has a model function property value of
Prediction,
verify that the following project properties are set:
-
Training Target Variable (for example,
lgd)
-
Class Target Level as
Interval
(SAS
Model Manager 12.1 or later)
-
Output Prediction Variable (for
example,
p_lgd
)
-
-
Right-click
Models and
select
Import fromLocal
Files.
Note: If the model already exists,
you can right-click the model name and select
Partial
Import to import an empty score.sas file, and then skip
to step 11.
For
more information, see Import Partial Models.
-
Navigate to the folder
on your computer that contains the component files for your model.
-
Select a classification
or prediction template from the
Choose a model template list.
-
Enter a text value in
the model
Name field.
-
Complete the template
fields. Drag the files from the left of the window to the corresponding
file property on the right. The following files are required:
-
Click
OK.
After SAS Model Manager processes the model import request, the new
model appears in the
Models folder of your
project's version.
-
Select the model in
the Project Tree, and set the model-specific properties. The value
for the
Score Code Type property must be
set to
DATA step.
-
Right-click the model,
and select
Set Model Output Mapping in order
to set the output variable mappings for the model. Click the list
in the
Models Variables column and select
the model output variable. Click
OK.
-
-
Before defining a performance
task, verify that the performance data set is registered in SAS Management
Console or that a libref has been defined for the performance data
set library using the
Edit Start-up Code window.
Make sure that the task contains the following variables:
-
Note: You must have the variable
columns in the table, but the values can be missing.
-
-
-
variables for characteristic analysis
-
Define a performance
task using the performance data set that contains the predicted values.
Also, make sure to clear the
Run scoring task option
for the
Data Processing Method section of
the
Define Performance Task wizard.
For more information,
see Run the Define Performance Task Wizard.