Personalization and privacy

How can businesses obtain critical e-commerce market data about Web customers while maintaining customer privacy? Assistant Professor Bamshad Mobasher's research project is tackling this hot e-commerce issue and may have found solutions through innovative techniques involving Web personalization and Web data mining.

Mobasher leads a research team that is developing automatic Web personalization software. The system obtains useful but anonymous information about Web site visitors by tracking their collective patterns of clicking through a site. Mobasher's group has developed Web usage mining techniques that allow site owners to gain deep knowledge about user patterns and behavior, even if personal or identifying information about users is not available.

"Customer privacy is a big problem with current systems," Mobasher explains. "Web sites currently collect customer data by requiring users to register personal information and interests or by tracking individual customer purchases. Some sites share cookies or other user information with vendors. Then they use this data to target their sales pitches or advertising.

"What we are developing is different," he says. "Our system collects customer data based on the clicking patterns of all customers who visit a site. It analyzes these navigation patterns and automatically generates targeted product offers and recommendations tailored to individual customers who are visiting the site. The system doesn't know any personal information about the

individual user, only the aggregate navigation and purchase patterns of all users"

The system also identifies the clicking patterns of non-customers versus customers, which can help site owners develop recommendations, such as discount coupon offers, that could turn browsers into buyers. It also can generate cross-sales.

In addition to maintaining customer privacy, tracking customer preferences through Web surfing patterns offers several benefits for e-commerce site owners, Mobasher says. "Data collected through user registration is subjective and prone to biases. It's also static-customers register their preferences once, but their tastes change over time. This system collects data objectively through the actions of all users and the data is more dynamic."

Mobasher, whose research focuses on data mining and intelligent computer agents, began work on the automatic Web personalization project in 1996 at the University of Minnesota. He brought the research to DePaul when he joined the CTI faculty last year. His project team includes Assistant Professor Craig Miller, four CTI students, and professors Jaideep Srivastava and Robert Cooley of the University of Minnesota.

Mobasher is preparing to launch a new center on Web data mining for e-commerce at DePaul, which will focus on research, development and interaction with industry. "Web data mining is becoming a very popular area of study, and there's a great deal of interesting and useful research that can be done in the field," he says.

More information about this research is available on the Web site: http://maya.cs.depaul.edu/~mobasher/personalization/ . Mobasher can be reached at mobasher@cs.depaul.edu .