Introducing New Mining Capabilities

by Anne Milley

SAS has a rich and proven history of providing an integrated set of analytical tools for a wide range of applications including forecasting, optimization, experimental design, quality improvement and predictive and descriptive modeling. In the summer of 2002, SAS Text Miner was first released to expand our customers’ ability to analyze not only structured data sources, but also to gain knowledge and insight from the vast deposits of textual data collected throughout corporate and research organizations.

Responding to increased demand to distill even greater value from data – both structured and unstructured – SAS is pleased to offer new analysis capabilities in SAS 9.1 for both Enterprise Miner and SAS Text Miner.

Enterprise Miner and Text Miner have both been redesigned for SAS 9.1 using an innovative Java client/SAS IOM server architecture to separate the mining computational server from the user interface work stations. This provides unprecedented flexibility for configuring an efficient installation that scales from a single-user system to a very large enterprise solution. Powerful servers may be dedicated to computing while end users move from office to home to remote site without losing access to mining projects or services. Mining processes can be run in parallel and scheduled in batch.

Batch processing was the No. 1 Enterprise Miner customer request for SAS 9.1. The optional Java Application Middleware enables clients to disconnect from a running diagram on the server or facilitate multiple users working with common SAS servers and a common SAS Metadata Server. Some process-intensive server tasks such as data sorting, summarization, variable-selection, and regression modeling have been rebuilt to distribute their work over multiple CPUs on the same system. All project information is stored on the SAS server for better wide-scale access and management throughout an organization. The user interface for both products continues to be based on the popular and productive process flow environment but is now built using Java Swing libraries for delivering more advanced graphics and visualization techniques with improved actions and controls. An easy-to-use Java Graphics wizard is also provided for developing customized graphs.

Other Enterprise Miner highlights include several new nodes and procedures, including Web path analysis, hierarchical associations, improved interactive decision tree training, autoneural networks, rule induction and more. Diagrams can be saved and imported as XML files for easy distribution to other data analysts. Complete compressed model results packages can also be created and registered to the metadata server for subsequent querying by data miners, business managers and data managers via the Web-based model repository. In addition to providing SAS, C and Java score code (including data preparation steps preceding modeling), model deployment through PMML will be available. An experimental Java API is also included for building customized Java applications that embed Enterprise Miner processes.

Other Text Miner highlights include additional language support, interactive graphics for linking significant concepts and hierarchical clustering output, the ability to analyze multiple text fields, Web crawling, additional parsing options and more. Since text mining, in general, is processor-intensive, improved performance also will be seen in Text Miner for SAS 9.1. Text Miner is fully integrated with Enterprise Miner, hence the distributable algorithms in Enterprise Miner also can be leveraged.

Collectively, these enhancements to SAS Enterprise Miner and Text Miner provide a more scalable and distributable set of tools for mining both structured and unstructured data sources to better address difficult business questions for even greater ROI.