A Network Diffusion Family of Ranking Methods, Stephen Devlin, USF

Friday, February 16, 2018 - 12:30
University of San Francisco Seminar Series in Analytics
San Francisco

**Meetups are streamed live from our facebook page at https://www.facebook.com/usfca.msds/. Coffee is provided at our 101 Howard location during the in-person seminar**

Speaker: Stephen Devlin

A Network Diffusion Family of Ranking Methods

Ranking items though paired comparisons is a problem with broad applications, especially in ranking teams or players in sports. We will present some of the most commonly used methods like the Markov method (a cousin of PageRank), and the methods of Colley and Massey (both of which were part of the BCS football rankings). We then introduce a parametric family of rankings using diffusion dynamics on the network of team connections, and show how this method can generalize, and interpolate between, several of the existing methods. We use NCAA DI college basketball data for examples and illustrations. Stephen Devlin is an Associate Professor of Mathematics at the University of San Francisco. He received his PhD at University of Maryland and was a CLE Moore Instructor at MIT before coming to USF.