Collaborative Filtering and the Netflix Prize

Abstract

Recommendation systems, and specifically collaborative filtering, are fairly new fields, which focus on automatically predicting how much a given user would value a given item. As more and more aspects of society are entering the digital realm, the value of such systems is increasing. Improvements to the underlying algorithms and concepts are still being discovered, and further research is being encouraged. Notably, Netflix recently issued a challenge to the collaborative filtering community, with a million-dollar prize offered for whomever could produce a significant improvement over their existing recommendation system. This paper explores the history of collaborative filtering, and many of the core concepts and methods. In addition, the progress of the winning teams from the Netflix Prize is tracked and their methods explored. Finally, there is a look ahead at potential challenges that those working in the field of collaborative filtering will have to overcome.