Collaborative Filtering: Netflix Prize

Speaker(s)

Arkadiusz Paterek

Affiliation

Uniwersytet Warszawski

Date

Dec. 14, 2006, 12:15 p.m.

Room

room 5870

Seminar

Seminar Algorithms

---

In this talk, I will present different algorithms for the collaborative filtering task. The task of collaborative filtering is to predict the preferences of a user, given preferences of other users. An example of a commercial collaborative filtering system is the movie recommendation algorithm Netflix's Cinematch, which predicts the customer's rating for a given movie. Recently, Netflix, the world's largest online DVD rental service, released to the public a large database of 17770 movies, rated by 480189 customers, and announced a $1M prize for 10% improvement in prediction accuracy over Cinematch on the dataset. I will give an overview of my solution - better than Cinematch by over 3%.

---