A tutorial at ECML PKDD 2015 about:

Collaborative Filtering with Binary, Positive-Only Data

By Bart Goethals (University of Antwerp), Kanishka Bhaduri (Netflix Inc.) and Koen Verstrepen (University of Antwerp)


Traditional recommender systems assume the availability of explicit ratings of items from users. However, in many applications this is not the case and only binary, positive-only user feedback is available in the form of likes on Facebook, items bought on Amazon, videos watched on Netflix, adds/links clicked on Google, tags assigned to a photo in Flickr etc. Recently, the number of publications on designing recommender systems that handle binary, positive-only feedback, is growing very fast. In this tutorial we discuss why collaborative filtering with binary, positive-only feedback is fundamentally different from collaborative filtering with rating data. We give an overview of the algorithms suitable for this task with an emphasis on surprising commonalities and key differences. Additionally, we provide extensive experimental comparisons among the most important algorithms. Finally, we discuss the role of these algorithms in the light of Netflix recommender system.


Part I: Introduction & Overview