What's New in SAS Model Manager 13.1

The SAS Model Manager Java client application has been replaced with a web-based interface that combines the managing of models, data sources, and workflows into one user interface. The application URL has been changed to be consistent across the SAS Enterprise Decision Management product offerings. The project control groups feature has been renamed portfolios, and the Models category has two separate subcategories for projects and portfolios. For more information, see Exploring the User Interface.

The Data Tables category enables you to create libraries and register tables to the SAS Metadata Repository. The tables can then be used as data sources when working with modeling projects. Tables that are already registered in the SAS Metadata Repository using the SAS Management Console can also be added to an existing library. For more information, see Managing Data Tables.

The SAS Model Manager Workflow Console has been merged with the new SAS Model Manager web-based application. You can now manage your workflows and perform tasks in the same user interface that is used to manage your modeling projects. For more information, see Managing Workflows.

The life cycle functionality has been deprecated and replaced with functionality that leverages SAS Workflow. Only migrated life cycles can be viewed. For more information, see View Life Cycle Status.

Support has been added for publishing models to the Hadoop Distributed File System and to the SAP HANA database. Published model files can also be removed from Hadoop, like the other publish destinations. However, published model files cannot currently be removed from an SAP HANA database. For more information, see Publishing Models to a Database.

From the Versions page within a project, you can add new versions, lock or unlock a version, and switch the displayed version. One or more versions can be active at one time, but only one can be the champion version. For more information, see Overview of Project Versions.

In addition to being able to execute a model retrain definition, scoring test, or performance definition, you can schedule either a one-time job or a recurrent job. You can have multiple scoring tests with different schedules. Each scoring test can have only one scheduled job. A retrain definition and a performance definition can have only one scheduled job at a time. For more information, see Schedule a Retrain Definition, Schedule a Scoring Test, and Schedule Performance Definitions.

When retraining models, you can specify to retrain a model if the dashboard project status is Alert or Warning. The dashboard project status is based on the project champion model. For more information, see Overview of Dashboard Reports.

When monitoring the performance of a champion or challenger model, you can specify static data sources or a library that contains data sources as part of the performance definition. An option has also been added so that you can generate the dashboard reports after performance monitoring has been completed. For more information, see Overview of Performance Monitoring.

The control group response rate, prior probability, and input table report options have been added to the Dynamic Lift report. The ability to select the input table and then select the variables to include in the summary data set has been added to the Training Summary Data Set report. For more information, see Overview of Model Comparison, Validation, and Summary Reports.

Documents can now be attached at the portfolio, project, and model object-level. All migrated attachments from the previous release that were stored at the organizational folder object-level, version object-level or in the Documents folder are now located on the Attachments page for the associated project. Comments can also be added at the portfolio, project, and model object-level. Comments that have been migrated and are associated with an existing workflow are available on the Comments page of a project. For more information, see Add Attachments and Add Comments.