The Lighter Side of Data Mining
By Linda L. Briggs
Can data analysis be fun? D'es predictive modeling have a lighter side? This spring, MBA students at the University of Washington will compete in a lighthearted “Crack the Case” contest to see who can do the best job analyzing a complex set of data. Using a graphical data analysis tool donated by Tableau Software, which is sponsoring the contest, the 13 teams of 2 to 3 students each will compete for HP notebook computers for the winning team.
A preliminary judging round will select three finalist teams. Those three will then present their case in front of a panel of judges that will include business executives from some high-profile Seattle-based firms, including Starbucks, Microsoft, and Internet traffic firm F5 Networks.
Behind the contest is Dr. Doug “Mac” MacLachlan, professor of marketing and international business at the University of Washington. MacLachlan teaches data mining principals, statistical algorithms and predictive modeling to his MBA students. These topics are so cutting-edge that he hasn’t yet found a suitable textbook.
“Increasingly, to get a sustainable competitive advantage in business, you need to approach decisions analytically,” MacLachlan says. “There’s so much data available that those companies that learn how to use it properly can get a leg up on competitors.”
As a judge, MacLachlan says he’ll be looking at the maturity with which students perform the task using Tableau. “Any data analysis can be confusing,” he says. “There are quirks to the [contest] data that have to be understood before going directly to the result.” Tableau created the simulated data based on real-world information, he says, and “there are some tricky features to it.”
Like many business intelligence products, Tableau aims to help users make sense of masses of data. It differs from other data analysis offerings largely in its simplicity of use and easy-to-grasp, visual presentation of data. The product can take massive amounts of numeric data and, with a minimum amount of user direction, produce colorful, useful charts, graphs, and tables.
MacLachlan decided to try out the Tableau software in his course on database marketing. “It seemed like interesting software to demonstrate how one might do fairly sophisticated kinds of drilling-down into a database,” he says. In class, he typically uses high-end statistical software for tasks like building models to forecast responses to an event like a direct mail campaign.
However, data mining tools are typically complex and costly. Although MacLachlan uses some products from companies like SPSS Inc. for his database marketing course, and Enterprise Miner from SAS for his doctoral seminar, he can’t always afford the tools he might like to use. For example, SPSS makes a data mining tool called Clementine, but “it’s just too expensive to buy for the students,” MacLachlan says.
One advantage of Tableau is its affordability. A license for classroom use of Tableau Standard, which connects to data in Microsoft Excel and Access, is $499. Tableau Professional, which connects to data in Microsoft SQL Server, Excel, and Access, along with MySQL and Oracle, is $899. Both prices include a year of software upgrades and unlimited technical support. Tableau will also work with faculty who teach courses around data analysis or data visualization. The vendor offers the product at no charge to qualifying instructors. (At the University of Washington, Tableau is letting MacLachlan use the product for three months in class at no charge.)
MacLachlan is evaluating his use of Tableau in class after this term. Since much of what he teaches is oriented toward predictive modeling, in which data is generated partly through experience with customers and partly through experimentation, the product isn’t an exact fit with the course. “My database marketing is oriented more toward statistical algorithms,” he says. “I wouldn’t orient a whole class around [Tableau], but it d'es have some pretty powerful features. You can quickly and powerfully drill down and get at features of the data, but you have to know where you’re looking.”
Learning data analysis techniques is powerful, MacLachlan says, because “students can go out and immediately become her'es in their companies. For the most part, companies haven’t figured out what to do about their data. I can show [students] how to grab the low-hanging fruit pretty quickly.”
Linda Briggs is a freelance writer based in San Diego, California.
