Fordham University
Department of Computer & Information Science
And
The Graduate Student Association
Present
 

Date:	Friday, April 27, 2007; 3:40pm
Location:	Lincoln Center, LL 816

Early work in predictive data mining did not address the complex circumstances in which models are built and applied. It was assumed that a fixed amount of training data were available-at no cost-and only simple objectives, namely predictive accuracy, were considered. Over time it became clear that these assumptions were unrealistic and that economic utility had to be considered during the three main stages of the data mining process: data acquisition, model construction, and model application. This led to work on active learning and cost-sensitive learning. However, most of this work factored in utility considerations in only one stage of the data mining process and did not consider interactions between the different stages. Recently, several colleagues and I proposed the term Utility-Based Data Mining (UBDM) to encompass all work that considers utility during the data mining process. Our goal in doing this is to encourage researchers and practitioners to pay proper attention to the role of utility and to consider how one can maximize utility throughout the entire data mining process. In this talk I will give a brief introduction to Utility-Based Data Mining and the subtopics it encompasses, discuss why UBDM is important, and review how existing research fits into this framework.

Bio:
Gary Weiss is an assistant professor in the Computer and Information Science Department at Fordham University. His current research interests include machine learning and data mining and the fundamental issues that arise when tackling complex, real-world problems. Over the past two years he has organized two workshops on Utility-Based Data Mining and is currently editing a journal special issue on this topic. Prior to coming to Fordham, he worked at Bell Labs and then AT&T Labs, spending his last five years at AT&T applying data mining methods to business and marketing problems. He received his B.S. from Cornell University, his M.S. from Stanford University, and his Ph.D. from Rutgers University.