July 24, 1997

Contact:
Leila Belkora
UIC Public Affairs
[email protected]
(312) 996-3457

At a recent conference on "data mining" at the University of Illinois at Chicago, experts from around the world witnessed a first in the use of the next generation of network communications.

"Data mining -- the automatic search for patterns, associations and changes in large databases -- is important for science and business," said Bob Grossman, conference organizer and UIC professor of mathematics, statistics and computer science. "This demonstration is important because it showed data mining can be done over a wide geographic area."

Potential applications of data mining include the detection of credit-card fraud and the improved understanding of health care data. Up to now, data mining has required that all of the data be in one location. However, scientific and corporate data are generally distributed over a wide geographic area. The data mining demonstrated at the conference used a state-of-the-art link between the University of Illinois at Chicago and the University of Pennsylvania.

The link set up at the "Managing and Mining Massive Data" conference July 12 to 15, sponsored by the National Science Foundation, relied on the very-high-speed Backbone Network Services provided by MCI under contract with the National Science Foundation. The network operates at speeds of 155 million bits of data per second -- or up to 100 times faster than the current Internet for certain applications.

In this demonstration, conference participants in Chicago not only interacted with their peers at the University of Pennsylvania through high-quality videoconferencing but simultaneously used the same communication channel to mine a scientific data set stored in Philadelphia. For this trial the demands on the network were about 18 times greater than the load carried by typical Internet connections available at universities, said Grossman.

"The data mining involved sustained 100-gigabyte (a gigabyte is one billion bytes) queries on more than a terabyte (one trillion bytes) of information that was distributed between two cities. That's much harder than analyzing the Library of Congress card catalog and the numbers from every phone book in the United States," said Grossman, who is also president of Magnify, Inc. in Chicago.

The data mining demonstration also achieved a milestone in the way information was exchanged. While the Asynchronous Transfer Mode, or ATM, communication protocol has been available for several years, most applications over new high-speed networks involve an older and less efficient protocol called TCP/IP, simply because that is what is available, according to Stuart Bailey, senior engineer at UIC's Laboratory for Advanced Computing. The experiment at the data mining conference demonstrated the full potential of the Asynchronous Transfer Mode over the high-speed network.

"This represents the first demonstration of high-performance data mining that I know of, and the first teleconferencing with so-called 'native ATM' over the very high speed Backbone Network Services," added Grossman.

The videoconferencing, data mining, and other demonstrations at the conference are part of the 1997 Terabyte Challenge, a year-long technology demonstration for data mining and data-intensive computing on massive data sets involving university and commercial scientists.

-UIC-

MEDIA CONTACT
Register for reporter access to contact details