The University of Arizona

Events & News


DateThursday, April 8, 2010
Time11:00 am
LocationGS 906
DetailsLight refreshments - 9th floor Atrium - 10:45 am
SpeakerYifan Hu
AffiliationAT&T Research Labs

What's Worth Watching on TV? -- a Collaborative Filtering Based Recommender for Implicit Feedback Datasets

Abstract: Interactive TV nowadays can offer many hundreds of channels. This makes finding interesting programs through channel flipping or via a program guide almost infeasible. A TV recommender system can improve customer experience through personalized recommendations. However, unlike the extensively researched instances with explicit feedback (e.g., Netflix
data), here there is no explicit feedback available. This is a typical
example of implicit feedback datasets, based on systems passively
tracking user behavior, such as purchase history, watching habits and
browsing activities. We do not have any direct input from the users
regarding their preferences. In particular, we lack substantial evidence
on which products consumer dislike. In this work we propose a new SVD
based model of implicit feedback datasets based on preference and
confidence. We also propose a solution procedure which scales linearly
with the data size. The algorithm is used successfully within a
recommender system for television shows. It outperforms well tuned
implementations of other known methods. We also offer a novel way to explain the recommendations via interactive visualization of the space of TV shows.