New Models and Methodologies for Group Decision Making, Rank Aggregation, Clustering and Data Mining

DANIEL J. EPSTEIN DEPARTMENT OF INDUSTRIAL AND SYSTEMS ENGINEERING SEMINARTitle: "New Models and Methodologies for Group Decision Making, Rank Aggregation,
Clustering and Data Mining"Guest Speaker: Dr. Dorit S. HochbaumHaas School of Business and Department of IE&OR, University of California, BerkeleyABSTRACT: We introduce models for problems of group decision making, aggregate ranking and clustering techniques for data mining. The problems are modeled as graph problems. One of these problems we call the equal paths problem. This problem as well as all problems studied here have convex objective function representing penalties for deviating from specified a-priori comparison/ranking beliefs. These problems are shown to be solvable in polynomial time using network flow techniques such as parametric cut and fractional multicommodity linear programming.One application of the aggregate ranking problem is to determine the ranking of sports teams based on the outcomes of games played. Current techniques are based on finding a maximum eigenvector. Our alternative model has a number of advantages including the ability to differentiate between games based on some measure of significance. Further, the problem is stated as a combinatorial graph problem. This problem is shown to be solved in polynomial time even with a convex objective function, using flow techniques.A closely related area that addresses various forms of rankings is data mining with applications to customer segmentation, patient diagnosis and assessment of bankruptcy risk. We demonstrate new models for these problems and how to solve them with flow techniques. Similarly, the models and solution methodology are applicable to multi-criteria decision making.MONDAY, DECEMBER 1, 2008, ANDRUS GERONTOLOGY BUILDING (GER) ROOM 309, 10:00–11:00 AM

Location: Ethel Percy Andrus Gerontology Center (GER) - 309

Audiences: Everyone Is Invited

Contact: Georgia Lum