Managing Models Using SAS Model Manager

You use a single interface, the SAS Model Manager client, to access all of your business modeling projects. The SAS Model Manager client presents projects in a tree structure, known as the Project Tree.

All models are stored in a central, secure model repository.

All project or model metadata is readily accessible through the SAS Model Manager client.

You can track the progress of your project’s version either by creating processes and activities using the SAS Model Manager Workflow Console or by creating milestones and tasks using a life cycle. You create custom processes and activities or milestones and tasks to meet your business requirements and to match your business processes.

Data tables that are registered in SAS Management Console or in an accessible SAS library can be used in SAS Model Manager.

The models that you import into SAS Model Manager can be SAS Enterprise Miner models, SAS/STAT linear models, SAS/ETS COUNTEG and SEVERITY models, models that you develop using SAS code, PMML models, or R models. You can create custom model templates for SAS code models so that SAS Model Manager knows exactly what files and metadata are associated with a model.

After you import candidate models, you can use SAS Model Manager to schedule and run scoring tasks to validate models.

SAS Model Manager has several reports that you can use to compare and assess candidate models. You can also write your own SAS reporting programs to assess candidate models and run them in SAS Model Manager. The aggregated reporting facility enables you to combine multiple reports into a single report.

After you choose a champion model, you can lock the model and its associated data for future reference or auditing by freezing the containing version.

SAS Model Manager uses the SAS Integration Technologies Publishing Framework to publish models to a channel.

You can flag challenger models and publish them to a production environment.

You can publish models to the SAS Metadata Repository, or you can publish the champion model and challenger models to a database for scoring using the SAS Scoring Accelerator.

You can monitor the performance of a champion model in a production environment by scheduling a performance monitoring task to run on a specific day and time, by executing the performance monitoring task using the SAS Model Manager window, or by using SAS Model Manager macros in a batch environment. The performance of challenger models can be monitored in a production environment using the SAS Model Manager window. After the data is collected, you can create a report that compares the performance of the champion and challenger models.

SAS Model Manager provides macro programs for you to run model registration to SAS Model Manager and scoring in a batch environment. Another macro registers SAS/STAT item store models and High-Performance Analytic models that were not created in SAS Enterprise Miner to the SAS Metadata Repository. If you create models using the COUNTREG or SEVERITY procedures, SAS Model Manager provides macros for you to generate score code for the models.

You can retrain models to respond to data or market changes.

SAS Model Manager provides dashboard reports for you to monitor the state of projects using performance monitoring reports, and enables you to view the reports in a web browser.

If your environment supports multiple SAS Application Servers, you can select the application server to execute scoring tasks, performance monitoring tasks, and model retrain tasks.

Using a query utility, you can look for models by name or identifier, or you can look for tasks.

Users in the Model Manager Administrator Users group ensure that all aspects of the modeling project are configured and in working order. Users in the Model Administrator group can perform all tasks within SAS Model Manager.

Users in the Model Manager Advanced Users group can perform some of the tasks that the Model Manager Administrator Users group can perform as well as all tasks that users in the Model Manager User group can perform.

Users in the Model Manager Users group can perform development, validation, reporting, and publishing tasks with some Write access limitations.