Before you begin a project,
you must plan your project resources. Here is a list of questions
to consider and conditions to meet for a modeling project:
-
After you know which users are
assigned to a project, a SAS Model Manager administrator must ensure
that the user is assigned to the appropriate SAS Model Manager user
group and role. For more information,
see SAS Model Manager User Groups, Roles, and Tasks and the
SAS Model Manager: Administrator's Guide.
-
How do you want to structure your
project in the Project Tree? A project is a subfolder of an organizational
folder. The Project Tree enables multiple levels of organizational
folders so that you can customize how you structure the Project Tree.
For example, your Project Tree could be similar to your business departmental
hierarchy or it could list individual project names. For more information,
see Organizing the Project Tree.
-
What models do you want to use
in the project? If the models were created using SAS Enterprise Miner,
SAS/STAT, or the
SAS/ETS procedures COUNTREG and SEVERITY, all model
components are available to SAS Model Manager when you import the
model. If your model is a SAS code model that is not contained in
a miningresult.spk file or a model that was created by third-party
software such as R, you must ensure that you have imported all of
the model component files. For more information,
see Import SAS Code Models and R Models Using Local Files and Import PMML Models.
-
How
do you want to define your project input and output variables? When
you create a project, you can import the variables using input and
output prototype tables, copy the variables from an existing champion
model, or define individual variables. If you use prototype tables
to define the project input and output variables, the tables must
be registered in the SAS Metadata Repository using SAS Management
Console or you must create a libref for files that are stored on a
local or network drive before you create the project. For more information,
see About Defining Project Input and Output Variables.
-
What method do you want to use
to track the progress of a version? The Workflow Console enables you
to track the progress of activities from the version level. A SAS
Model Manager administrator can create a workflow and associate it
with a version. You can also use the life cycle feature to track the
life cycle of a model at the version level.
-
If
you decide to use the workflow process to track the progress of activities
for a version, you do not need to use the life cycle feature to monitor
the progress of milestones and tasks.
For more information,
see Overview of Workflow Console.
-
If you decide to use the life cycle
feature to monitor the progress of your version, you must plan your
milestones and the tasks for each milestone before you can create
a version for a project. When you have that information, you then
create a life cycle template. The life cycle template enables you
to assign users to complete projects and to monitor the progress of
your project. For more information,
see Creating Life Cycle Templates.
-
You might have project documents
that you would like to access from SAS Model Manager. SAS Model Manager
enables you to attach documents to a
Documents folder
in the Project Tree. You can view these documents in SAS Model Manager
only. For more information,
see Associate Documents with a Folder.
-
SAS Model Manager provides several
reports that you can use to help you assess candidate models. You
can review the types of reports that are available and plan for which
reports you want to use. Your plans might also include a custom report
that you can run in SAS Model Manager. For more information,
see Validating Models Using Reports and
Validating Models Using User Reports.
-
When you publish a project champion
model to the SAS Metadata Repository, you must specify a folder to
which you can publish the project champion model. You might need to
create a folder in the SAS Metadata Repository, if one does not already
exist. For more information,
see Publish Models to the SAS Metadata Repository.
-
After your champion model is in
a production environment, you can monitor the performance of the model
in SAS Model Manager using your organization's operational data.
If you use SAS Model Manager to define and execute performance tasks,
you must first prepare performance tables using the operational data
and add them as a SAS Model Manager performance data source. For more
information,
see Creating a Performance Table.
-
When you run performance monitoring
reports, you can set up performance index alert and warning conditions
to notify users if conditions exceed the indexes. For more information,
see Performance Index Warnings and Alerts.