SAS launches model manager product

Finding itself in the unusual position of playing catch-up, SAS announced SAS Model Manager today at its annual M2006 Data Mining Conference in Las Vegas.

Rival analytic firms SPSS and Teradata, a division of NCR, announced similar lifecycle management for analytic models products earlier this year.

All the products provide governance for the modeling lifecycle from creation to optimization to tweaking the model through production deployment, according to James Kobielus, principal analyst at Current Analysis. Neither IBM nor Oracle has a model managing tool to date, added Kobielus.

One of the major headaches SAS Model Manager will resolve, according to Mary Crissey, analytical marketing manager for SAS, is version control.

“It identifies the draft version from the production model,” said Crissey.

Other features include audit logs to track when a model was deployed and used or upgraded by whom and the ability to take live feeds for “scoring.”

Scoring evaluates to what extent a model has predicted outcomes correctly based on the current or new data coming in.

Other key features include the ability to set a default version for a particular project and attach supporting documentation in  Microsoft Word, Excel, and HTML formats,

However, the SAS tool is at this point proprietary and works only with a SAS modeling environment while both SPSS and Teradata can import and manage models created by other vendor tools using using PMML (Predictive Model Markup Language).

“Considering most organizations have different data mining tools, it makes sense to have one governance environment that can manage them all,” said Kobielus.

Crissey at SAS agreed and added that SAS is working on a PMML feature that will be available in the next version of SAS expected in the summer of 2007.

SAS Model Manager will ship in November.