After nearly three years, Netflix announced the winner to its Netflix Prize yesterday. The million dollar prize goes to “BellKor’s Pragmatic Chaos”, a team of seven engineers, statisticians and researchers from Austria, Canada, Israel and the U.S.
More than 51,000 people worldwide competed for the $1 million Netflix prize for three years, trying to improve the accuracy for predicting what movie a person may like to watch based on what movies he liked in the past and what movies people with similar tastes liked in the past. Netflix said that the winning solution will make the movie predictions twice as good as before. 0
The BellKor’s Pragmatic Chaos team was formed and collaborated over the Internet. The team met face-to-face for the first time before the award announcement news conference at New York. Netflix CEO Reed Hastings and Chief Product Officer Neil Hunt announced the winners, handed them a check of $1 million and an Olympic-size “gold” medal for each of the team members. 0
The BellKor’s Pragmatic Chaos team consists of:
Bob Bell and Chris Volinsky, members of the Statistics Research Department at AT&T Research, and winner of the 2007 and 2008 Progress Prizes as members of team BellKor.
Andreas Toscher and Michael Jahrer, machine learning researchers and founders of commendo research and consulting in Austria. They won the 2008 Progress Prize as members of the team Big Chaos.
Electrical engineer Martin Piotte and software engineer Martin Chabbert of Montreal, and founders of Pragmatic Theory.
Yehuda Koren, Senior Research Scientist at Yahoo! Research Israel, and winner of the 2007 and 2008 Progress Prizes as a member of team BellKor.
Bellkor only edged out the previous first ranked leader “The Ensemble” at the last few minutes of the contest.
Bellkor used matrix factorization models for their algorithm. This winning shows that matrix factorization models are superior to classic nearest-neighbor techniques for producing product recommendations, allowing the incorporation of additional information such as implicit feedback, temporal effects, and confidence levels.
Netflix now has announced Netflix Prize 2: It will focus more on helping customers early in their experience with Netflix, drawing on many more sources of data to try to find just the right movies.