New version helps to operationalize insights, streamline analytical processes, meet governance mandates
Analytical models are high-value assets that improve decisions and business outcomes. But, as models degrade, so do the conclusions they support. The new release of SAS® Model Manager, from the leader in business analytics, makes it easier and quicker to register, validate, deploy, monitor performance and retrain models. Direct marketing, fraud detection, credit-risk and warranty applications, for example, employ hundreds of models, each requiring multiple deployment and management tasks and processes. SAS Model Manager closes the gap between model development and deployment, extending the usefulness of analytical models and ensuring timely and optimal results.
“Using SAS Model Manger as the analytic deployment hub, models can be deployed in a SAS environment taking full advantage of the underlying SAS platform and its metadata management capabilities,” according to Victory Index report author Fern Halper, Partner at Hurwitz & Associates. “Models can also be deployed directly via a number of in-database partnerships, using the SAS Scoring Accelerator.”
Import, embed, share models from SAS and more
Data scientists and analysts who create models using different tools enjoy a common framework to evaluate and report on how each model compares against test data. They can quickly determine which model to put into production. In addition to importing SAS Enterprise Miner models, SAS Model Manager provides a central, secure repository for registering, comparing, reporting, scoring and monitoring SAS/STAT, SAS High-Performance Analytics Server, open-source R, and PMML-compliant models. It deploys the models against operational applications. Models can also be deployed using SAS Data Integration Server and SAS Real-Time Decision Manager.
“Customers can use diverse tools, with SAS as the analytical hub,” said Tapan Patel, Data Mining and Predictive Analytics Product Marketing Manager at SAS. “In response to customer requests, SAS Model Manager now helps users integrate methods from other tools and effectively manage all their models in one repository. SAS provides customers more options to derive value from data, using proven and validated analytical methods in SAS and importing emerging methods not yet implemented in SAS. Today’s enhancements demonstrate SAS’ support for standards and interoperability.”
Cross-functional analytics management
Analytics, validation, and IT teams use SAS Model Manager collaboratively to provide complete lifecycle management and governance of analytic models.
- Data scientists and other analytic professionals can validate, deploy and monitor performance of their models. When model performance degrades, they can take quick action. The new Web-based SAS Model Manager Workflow Console streamlines the model approval process and improves productivity.
- Risk officers and compliance executives can benefit from extensive registration, validation, version control and documentation for analytical models. This is critical for highly regulated industries such as banking. The new release provides validation reports for Basel II risk models, including probability of default and loss given default. It will help gain transparency and answer regulator inquiries on demand.
- IT teams can automate model deployment into the production environment, publishing and integrating models into operational applications to deliver faster results. Scoring and performance monitoring can be done on a database appliance (Teradata or EMC Greenplum), configured for use with SAS High-Performance Analytics Server.