After you create a project,
you import models into a project version on the
Models page.
A project can contain multiple versions. You can see only the models
for the selected version on the
Models page. After model evaluation, you set one of the
candidate models as the
champion model.
There are many methods
of importing your SAS models into your project version:
Keep the following details
in mind:
-
Scorecard models can be imported
using the SAS Code Models local files method and the SAS Model Package
File import method.
-
HPFOREST procedure models can be
imported using the SAS Metadata Repository import and the SAS Model
Package File import. You cannot import PROC HPFOREST models using
local files.
-
High-Performance
analytics models that are not created with SAS Enterprise Miner can
be registered to the SAS Metadata Repository using the %AA_Model_Register.
These models can then be imported to SAS Model Manager by importing the models from
the SAS Metadata Repository from a SAS
model package file.
-
Before you can import COUNTREG procedure and SEVERITY procedure models, you must create
the
model score code using the %MM_Countreg_Create_Scorecode macro and the %MM_Severity_Create_Scorecode
macro. After the
score code is generated, you can use the %MM_Model_Register macro or the local files method
to import these models. For more information about the types of model component tables,
see
Generating Score Code for COUNTREG Procedure Models.
-
SAS Model Manager cannot
publish models to a database whose Score Code Type model
property is set to SAS Program, PMML, Analytic
store, or DS2.
-
Model component table variable names must start with a letter or underscore, and
can contain letters, the underscore ( _ ), the hyphen ( - ), and the period ( . ).
Variables with
special characters can be used only when the administrator has set the
Valid Variable Name option to
Yes in the SAS Management Console or set the variable from
start-up code. For more information, see the
SAS Model Manager: Administrator’s Guide.
CAUTION:
Unexpected
results might occur if you import a model that was previously exported
using SAS Model Manager.
A best practice is
to import models that were not previously exported by SAS Model Manager.