SAS Software Takes Top Medals in KDD Data-Mining Competition
ACTION ITEMS

SAS Institute Inc.'s Enterprise Miner software and a product built on SAS software's analytic toolset have won silver and gold medals, respectively, in the 1998 KDD Cup, a data-mining competition sponsored by the 4th-international Conference on Knowledge Discovery and Data Mining. The Institute, the only enterprise-wide vendor to medal in the contest, and two other companies prevailed over 54 other contestants.

Companies were challenged to build fundraising net-return prediction models for a not-for-profit organization serving U.S. veterans. With an in-house database of more than 13 million donors, it is one of the nation's largest direct-mail fundraisers. Competitors "mined" donor data to help the organization maximize revenue from future mailings to donors.

The top two models were nearly identical, one being submitted by Urban Science Applications Inc., whose data-mining tool is built on SAS software analytics and application-development tools; the other by Enterprise Miner software. For Enterprise Miner software, unique in its ability to apply and compare multiple modeling techniques, the contest showcased just one of a wide range of business problems it solves daily for major banks, catalog retailers, department stores, telecommunications companies, and more.

"These wins confirm SAS Institute's leadership in data mining and analytic solutions," said Mark Brown, SAS Institute's program manager for data mining. "SAS Institute, the only enterprise vendor in the KDD Cup competition, is proud to have won top honors in this test of database-marketing prowess.

"At cross-industry sites around the world, our software is the tool of choice for many other data-mining applications as well. The KDD medal validates SAS Institute as a premier solution provider via Enterprise Miner software, which automates churn analysis, fraud detection, cross-selling, upselling, and much more. It also illustrates the Institute's value as a data-mining tool provider."

Leading the Institute's competition team were data-mining developers Warren Sarle and Padraic Neville, supported by the Institute's data mining testing and marketing team. Neville accepted the KDD Cup silver medal on behalf of SAS Institute.

Enterprise Miner software, now available on Windows NT and UNIX platforms, offers a common interface for both quantitative experts and business analysts. For their IT counterparts, it offers the most complete and integrated data-mining solution, able to access all corporate data no matter where it is or in what form.

Enterprise Miner software provided the Institute team many options for analyzing the data. Techniques considered include simple regression; two-stage regression with gamma error in the second stage; neural networks with several different architectures; plain and boosted naive bayes; plain, bagged and boosted decision trees; discriminant analysis; and ensemble models.

"Our final model was an ensemble model, which combined a linear regression, a two-stage regression, and a neural network," Sarle said.

SAS Institute's data-mining expertise is supported by more than two decades as a recognized leader in data-analysis R&D. SAS Institute consultants offer an integrated data-mining solution that builds on an end-to-end data-warehouse infrastructure. Consultants provide tools and software applications that incorporate the Institute's step-by-step SEMMA approach for understanding large quantities of corporate data.

SAS software and services are used at more than 31,000 business, government, and university sites in over 120 countries. Founded in 1976, SAS Institute is the world's largest privately held software company with annual revenue of $750 million.

For more information, see http://www.sas.com