【Topic】Predictive Data Mining under Asymmetric Costs

【Speaker】Huimin Zhao

【Time】2008-6-25 10:30-12:00

【Venue】Room 453, Weilun Building

【Language】English/Chinese

【Organizer】Department of Management Science and Engineering

【Background Information】

Predictive Data Mining under Asymmetric Costs

In real-world predictive data mining (classification or regression) problems, different types of prediction errors often incur asymmetric costs. Classifying a problematic bank as a healthy one, thus allowing the problems go unnoticed, has much more serious consequences than doing the reverse. Under-forecasting the loan loss amount of a bank, thus leading the bank to have insufficient reserves, is much more costly than over-forecasting the same amount. Such practical problems demand cost-sensitive learning methods that attempt to minimize the overall misprediction cost rather than simple performance measures like error rate and squared error. In this talk, I tackle the following questions related to cost-sensitive classification. What approaches can be taken to make basic learning methods cost-sensitive? What implications do these approaches have on different types of learning methods? In practical situations where the costs cannot be precisely pinned down, how should tools be designed to help decision makers make tradeoffs on different prediction errors? I will also present some preliminary work on tuning regression models to make cost-sensitive predictions.

Huimin Zhao is an Assistant Professor of Management Information Systems at the Sheldon B. Lubar School of Business, University of Wisconsin-Milwaukee. He earned his Ph.D. in Management Information Systems from the University of Arizona, USA, and B.E. and M.E. in Automation from Tsinghua University, China. His current research interests are in the areas of data mining and medical informatics. His research has been published in several journals, including Communications of the ACM, IEEE Transactions on Knowledge and Data Engineering, IEEE Transactions on Systems, Man, and Cybernetics, Information Systems, Data and Knowledge Engineering, and Journal of Management Information Systems. He is a member of IEEE, AIS, IRMA, and INFORMS Information Systems Society. He serves on the editorial review board of the Journal of Database Management.