SAS® Model Manager is getting an update soon to release 13.1 (I last blogged about Model Manager 3.1). The vision of SAS Model Manager going forward is to streamline the integration of predictive modeling into the overall environment, make it easier to operationalize analytical models, expand the model portfolio management capabilities and improve governance and monitoring of large numbers of models.
The new release will be standardized on the web-based application framework, making SAS Model Manager 100% browser-based – importing models, setting up champion-challenger, viewing performance etc. This web interface also makes these capabilities easier to access from within the SAS Decision Manager environment (reviewed here which includes the SAS Model Manager functionality).
In addition to bringing the existing functionality into the new web interface, , some new capabilities have been added. For instance, from within this new web-based SAS Model Manager interface it is now possible to explore and do calculations against the data defined in SAS Metadata Manager. This allows a SAS Model Manager user to view and understand this data in the context of viewing and managing the models registered with SAS Model Manager.
In addition, the SAS workflow engine is now completely integrated into the SAS Model Manager environment for those customers applying the workflow engine and its associated templates to the model management process. This allows tracking, assignment, new workflows etc. to be managed in context. Templates are provided for the workflow engine to handle standard analytic model management tasks like champion/challenger and import models.
Deployment from SAS Model Manager has been extended into the SAS Decision Management environment and the SAS Scoring Accelerator (which will support Hadoop and SAP HANA by the time the 13.1 version is released). This allows SAS Enterprise Miner and SAS/STAT models to be deployed directly into these environments from Model Manager.
Monitoring and retraining of models can be increasingly automated in SAS Model Manager with this release. Specific performance reports can be set up to run on a recurring schedule to update performance dashboards. A set of metrics and thresholds can be defined to trigger automated re-training of models or manual workflow tasks to prompt a user to retrain. Models can already be grouped into projects and now these projects can be grouped also into portfolios. Once grouped into projects or portfolios the models are managed as sets. This allows things like performance reports to be executed across all the models in the portfolio or variables to be added to all the models. When portfolios are created a wizard runs to allow large groups of models to be imported in one go. Once created portfolios and projects are managed directly, hiding the individual models and reducing the complexity of managing very large numbers of models.
You can get more information on SAS Model Manager here and SAS is one of the vendors in our Decision Management Systems Platform Technologies Report.