SAS^® Scoring Accelerator

High-performance model scoring and deployment with faster time to results

Overview
Benefits
Features
System Requirements

Fast answers with complete accuracy. Unfortunately, most organizations simply can't aggregate, analyze and process large volumes of data quickly enough to support critical decisions that must be made in hours, minutes or seconds.

Conventional model scoring requires the transfer of data from databases back to SAS. Scores must then be bulk loaded back to the database. This excessive data movement and latency creates inconsistencies and problems with data governance. Slow analytical processing results in model scoring delays. Ultimately, the inability to capture timely insights from growing volumes of data means missed opportunities.

SAS Scoring Accelerator automates model-scoring processes within the database so you can improve model scoring performance and achieve faster time to results. SAS Scoring Accelerator enables customers to translate scoring models created in SAS® Enterprise Miner™ or SAS/STAT® into database-specific functions to be deployed and then executed directly within the database environment.

Benefits

Achieve higher model-scoring performance and faster time to results.
Enhance the productivity of your data mining and IT groups.
Reduce data movement and latency, and streamline analytic deployment.
Better manage, provision and govern data.
Make better use of IT resources and reduce costs.

SAS^® format library
SAS^® Enterprise Miner™ Score Export node
Register SAS/STAT^® linear models to SAS^® Model Manager
SAS^® Scoring Accelerator publishing client
Integrated environment for tracking and monitoring model performance over time
Support for multiple databases

Provides timely and accurate insights needed for evidence-based decisions. SAS Scoring Accelerator simplifies and hastens the deployment of predictive models by publishing them as database-specific functions that are executed on all available database resources. You can reduce model scoring time from hours to minutes.
Tracks and monitors model performance over time to eliminate model decay. SAS Model Manager, an analytic model management and deployment environment, is fully integrated with SAS Scoring Accelerator to further streamline the registration and validation of SAS models in the database.
Delivers the best solution for successful organizations. SAS has partnered with several vendors to deliver the best analytics and data warehousing technologies – optimized and integrated for added value.

Benefits

Achieve higher model-scoring performance and faster time to results. Using in-database processing, SAS Scoring Accelerator eliminates the need to move data between SAS and the database for scoring purposes, further reducing the cost, complexity and latency of the scoring process. The performance of the entire modeling process is improved, enabling faster predictive results and competitive advantage.
Enhance the productivity of your data mining and IT groups. Rather than using manual transformation and scoring steps, SAS Scoring Accelerator allows IT to automate model deployment processes. Faster deployment frees the analytics team to focus on new projects. Organizations incur fewer labor costs by avoiding manual scoring because that reduces the need to revalidate code for scoring purposes.
Reduce data movement and latency, and streamline analytic deployment. In-database scoring reduces data movement and replication to streamline analytic processing. This helps organizations deliver analytics quickly while maintaining data reliability and integrity.
Better manage, provision and govern data. Reducing data movement and consolidating data in a data warehouse addresses data governance and helps ensure regulatory compliance. The common security, auditing and administration features of the database help ensure that regulated data is accessed properly. In addition, by running model-scoring steps inside the database, programs can be executed in such a way that enables detail data to be summarized and analyzed without crossing the regulatory boundary created in the database.
Make better use of IT resources and reduce costs. Proper integration and efficient use of investments in data warehouse platforms, analytic software and processes result in the delivery of consistent, reliable and timely insights to decision makers. Providing operational insights in an efficient manner yields faster time to value at a reduced cost.

Features

SAS^® format library

File that is deployed once to the database system.

SAS^® Enterprise Miner™ Score Export node

Automatically included in your SAS Enterprise Miner installation for adding to your process flow diagram analyses.
Exports the model-scoring logic, including metadata about the required input and output variables.
Generates sample training and a scoring data set for testing.

Register SAS/STAT^® linear models to SAS^® Model Manager

SAS Model Manager is required to register and publish linear models created using SAS/STAT into SAS Scoring Accelerator.
In-database scoring for SAS/STAT is supported for EMC Greenplum, IBM DB2, IBM Netezza, Oracle and Teradata.
Support is available for a select set of predictive models created using the following procedures: LOGISTIC, GENMOD, REG, GLMSELECT, GLM, GLIMMIX and MIXED.

SAS^® Scoring Accelerator publishing client

Automatically translates and publishes the model as a scoring function or SAS DS2 program inside the database.
Generates a script of database commands for registering the scoring function or SAS DS2 program inside the database.
Scoring functions or SAS DS2 programs are available for use in any SQL expression wherever database-specific built-in functions are typically used.
Publishes model as protected or unprotected.
Supports SAS intrinsic and user-defined formats.
Supports a robust class of SAS Enterprise Miner predictive and descriptive models, including the preliminary transformation layer (such as data imputations).
Provides variable binning and reduction.

Integrated environment for tracking and monitoring model performance over time

Fully integrated with SAS Model Manager (required for linear models created using SAS/STAT and optional for models creating using SAS Enterprise Miner), an analytic model management and deployment environment, to further streamline the registration and validation of vendor-defined functions (VDFs) in the database.

Support for multiple databases

Aster Database (only supports SAS Enterprise Miner models).
EMC Greenplum Database.
IBM DB2.
IBM Netezza 1000.
Oracle 11gR2 (single node, RAC and on Exadata Appliance).
Teradata Enterprise Data Warehouse (EDW).

System Requirements