MasPar reborn as NeoVista

MasPar Computer Corp., which specializes in building high-performance parallel-processing solutions in the federal market, is reinventing itself this week as NeoVista Solutions Inc. and bringing to market a suite of high-end data-mining tools for open-systems platforms.

In shedding its old name, NeoVista also sheds dependence on its proprietary systems, relying instead on the emerging capabilities of commercially available parallel-processing hardware and relational database software.

NeoVista, however, retains the expertise it gained in Defense Department and intelligence agencies in pattern recognition technology - the basis of data mining; this expertise can now be applied to more mainstream decision-support applications.

"We have been selling pattern recognition applications in things as [varied] as fingerprint recognition and radar sonar applications," said John Harte, president and chief executive officer at NeoVista, Cupertino, Calif. "Out of that, we have come to a good understanding of what large database pattern recognition applications are all about."

Decision Series Due This Year

Later this year, NeoVista will begin shipping Decision Series, an integrated suite of scalable data-mining tools that work with database software from Informix Software Inc., Oracle Corp. and Sybase Inc. and run on parallel-processing platforms from Hewlett-Packard Co., Digital Equipment Corp. and Sun Microsystems Inc.

Like other data-mining tools, Decision Series will allow users to run complex queries against large databases or data warehouses. But unlike traditional queries, data mining frequently involves searching databases for hidden patterns. In many cases, users are testing a hypothesis about a trend. However, users often take what is called a bottom-up approach with queries, searching for hidden patterns and answers to unasked questions.

Decision Series provides data-mining engines for neural networks, association rules, and genetic and clustering algorithms. It also includes software called DecisionAccess, which translates data back and forth between the engines and the database platforms.

"At this point, NeoVista is the first to market with the planned complete set of products," said Bruce Love, research director for applications of technology at The Gartner Group. NeoVista also appears to be offering more scalability than other tools providers, Love said.

NeoVista also has an edge in this market because its software can handle the level of detailed data required for that bottom-up approach, said Judson Groshowng, NeoVista's vice president of marketing. Ideally, users can use data-mining results to feed new questions into the data warehouse.

But that is not possible with less-scalable tools, Groshowng said. Once databases cross a certain threshold, most tools must rely on working with data summaries rather than the data itself, which cannot be fed back into the data warehouse. "As the amount of material gets larger and larger, the ability to understand the structure of the data gets harder and harder to find," Groshowng said.

Detailed data is a key concern of the Army and Air Force Exchange Services (AAFES), which is one of the first organizations evaluating NeoVista's software. AAFES runs 17,000 retail operations at DOD bases worldwide and has been using data mining to sort through operations results for nearly a decade, said Denise Barnhart, chief of AAFES' Corporate Analysis Division, Dallas.

But "when you get down to the item level and you are talking 600 items per store, it's huge, and you start losing resolution," Barnhart said.

Although Decision Series will appeal strongly to such commercial operations, NeoVista expects a lot of interest from the Social Security Administration, Internal Revenue Service and law enforcement agencies.

"We are looking at patterns of behavior where there is a tremendous amount of data and lots of people involved," Harte said. "We've been doing it for years with" the National Security Agency and the CIA.