Charles Elkan is a data mining expert, a computer science professor at the Jacobs School of Engineering at UC San Diego, and a consultant/judge for the Netflix Prize. The $1 million prize went today to BellKor’s Pragmatic Chaos.
Professor Elkan attended the prize ceremony in his capactiy as a judge for the competition. Below are a few insights from Elkan regarding the winning team's approach to movie recommendations. A short news story about Elkan's participation in the Netflix Prize is on the news section of the Jacobs School of Engineering news site.
"The success of the winning team [BellKor’s Pragmatic Chaos] is based on groundbreaking basic research stimulated by the contest. One important idea is to use hundreds of variations of multiple methods, and then to take a sophisticated average of their predictions. This way, the errors of the different methods can partially cancel out. A second important idea is to discover from the data factors that are relevant both in describing movies and in determining the preferences of viewers. These factors are mined from the data to be maximally predictive, as opposed to being programmed in by human experts. It turns out that mined factors are more useful than any human intuitions for this task. A third important idea is to model the subtle ways in which ratings from users change both day to day and over the long term," said Charles Elkan, computer science professor from the UC San Diego Jacobs School of Engineering
Journalists: Charles Elkan is available for comment regarding the Netflix Prize. He can provide clear insights and explanations regarding the winning strategies and how this work ties into larger data mining trends in both business and academia. Contact me (dbkane AT ucsd DOT edt) and I'll put you in touch with professor Charles Elkan.
Professor Elkan attended the prize ceremony in his capactiy as a judge for the competition. Below are a few insights from Elkan regarding the winning team's approach to movie recommendations. A short news story about Elkan's participation in the Netflix Prize is on the news section of the Jacobs School of Engineering news site.
"The success of the winning team [BellKor’s Pragmatic Chaos] is based on groundbreaking basic research stimulated by the contest. One important idea is to use hundreds of variations of multiple methods, and then to take a sophisticated average of their predictions. This way, the errors of the different methods can partially cancel out. A second important idea is to discover from the data factors that are relevant both in describing movies and in determining the preferences of viewers. These factors are mined from the data to be maximally predictive, as opposed to being programmed in by human experts. It turns out that mined factors are more useful than any human intuitions for this task. A third important idea is to model the subtle ways in which ratings from users change both day to day and over the long term," said Charles Elkan, computer science professor from the UC San Diego Jacobs School of Engineering
Journalists: Charles Elkan is available for comment regarding the Netflix Prize. He can provide clear insights and explanations regarding the winning strategies and how this work ties into larger data mining trends in both business and academia. Contact me (dbkane AT ucsd DOT edt) and I'll put you in touch with professor Charles Elkan.
The official Netflix press release is here.
Insightful story from Communications of the ACM by Marina Krakovsky. This story quotes Charles Elkan.
The Wall Street Journal story by Marisa Taylor mentions Charles Elkan as a judge.
