From: technews <technews@ou.edu>
To: "'it-fyi@listserv.ou.edu'" <it-fyi@lists.ou.edu>
Subject: it-fyi: Business Students Learn to Mine Data (Chron of Higher Ed)
Date: Fri, 5 Nov 1999 10:19:37 -0600
At Business Schools, Students Learn to Mine Data on Consumer Behavior
By FLORENCE OLSEN
Courses that you thought only computer-science majors could enjoy have
become favorites among M.B.A. students at business schools, where high-tech
methods for studying data about consumers are changing the way marketing and
management courses are taught.
At some of the schools, waiting lists for the courses -- with names like
"Decision Support," "Marketing-Intelligence Management," and "Advanced Data
Analysis for Decision-Making" -- have forced marketing and management
departments to expand the number of sections they offer, even though the
hardware and software required for the courses are expensive.
The new courses teach students about "data mining" -- techniques of using
computers to dig through huge data sets in search of veins of useful
information. The methods can be applied to all kinds of data, but what
interests businesses -- and business students -- is the possibility of
extracting information about customer behavior from data collected from
e-commerce Web servers, checkout-line scanners, credit-card terminals, and
other checkpoints of the consumer society.
"We're trying to estimate student demand to make sure we have enough seats
for the students next semester," says Raymond R. Burke, a professor of
business administration at Indiana University at Bloomington's Kelley School
of Business.
Mr. Burke devotes at least half of his "Applied Marketing Research" course
to teaching data mining. Sometimes, he says, only data mining can help
companies, governments, or institutions make sense of the huge amounts of
data their computers collect. Businesses used to gather market-research data
on a "one-shot basis," from focus groups, surveys, or test markets, says Mr.
Burke. Now they collect information on consumers every day.
"The neat thing about data mining is you can throw in the kitchen sink,"
says Arthur E. Weston, a recent graduate of the Kelley School, who landed a
job as a data modeler in the data-mining division of a Chicago bank.
Mr. Weston, an assistant vice-president of the bank, uses data bases
containing 500,000 to one million records each for his market research. He
analyzes 300 to 500 variables at a time, using a computer powerful enough to
run a mid-size business.
With their retail-accounts data, bankers like Mr. Weston are able to build
models that predict which of their customers are likely to purchase new
certificates of deposit or other financial products, and which are not. The
banks use such information to refine their marketing campaigns, avoiding
wasteful spending on unlikely buyers.
Data mining is so new, Mr. Weston says, that his business-school experience
didn't teach him how to avoid its pitfalls, such as "false artifacts" and
other potentially costly mistakes. False artifacts are misreadings of data
that could lead a company to make bad marketing decisions.
Mr. Weston says false artifacts can be avoided by analyzing historical
marketing data about individual consumers, collected before they bought a
given product. But that strategy, he says, is not usually taught in business
school, where data mining is new and where data bases that can serve as
teaching tools are both hard to come by and impractical to use.
Businesses now rely on data mining for a variety of tasks, from detecting
credit-card fraud to discerning hidden predictors of charitable giving to
colleges.
In data-mining courses, business-school faculty members and students are
taking advantage of new software packages that offer users simplicity, says
Indiana University's Mr. Burke. He adds, "You don't need to know a lot about
statistics to use them."
Data-modeling algorithms, the computational software used in data mining,
behave differently from traditional statistical algorithms, says Haim
Mendelson, a professor of information systems and management who is
co-director of the Center for the Study of Electronic Commerce at the
Stanford Graduate School of Business. The techniques themselves come from a
field of computer science called machine learning, which borrows ideas from
statistics, complexity theory, and cryptography.
Despite its origins in machine learning, data mining "is not a super-exact
science," says Nenad Jukic, an assistant professor of information systems at
Loyola University of Chicago. He is also coordinator of the university's
certificate program in data warehousing and business intelligence.
"You're extracting previously unknown information," Mr. Jukic says. "But you
still have to have an excellent marketing person to interpret it and tell
you if it means anything."
At least five techniques, including neural-network modeling, which mimics
the activity of the human brain, are a part of data mining, says Donald Lee
Harnett, a professor of decision sciences at Indiana's Kelley School.
The institutions that are using data mining have "tons of data" and can
afford the computers and the software; the latter alone can run as high as
$80,000.
"We're interested in solving problems using computers," he says, "and that's
where data mining fits in."
Ajay Vinze, an associate professor of accountancy and information management
at Arizona State University, says that 45 faculty members and administrators
attended a two-hour introductory workshop on data mining that he held last
month.
His M.B.A. course on "Decision-Support Systems" is usually overbooked, and
he says he'd like to interest other faculty members in teaching data-mining
methods in their courses. Many types of institutions "are sitting on a lot
of data," he says, "and trying to figure out what in the world to do with
it."
Mr. Harnett says colleges and universities themselves could benefit from
data-mining methods to analyze student or alumni information. "My guess is
that they haven't gotten around to doing it," he says. "They're too busy
trying to do a hundred other things."
_______________________________________________________________
Copyright 1999 by The Chronicle of Higher Education