The Netflix Challenge
The Advertising Age article that was required reading for this week’s COM 529 class talked about the Cinematch application in Netflix. Netflix tracks your rental history and allows you to rate movies, then Cinematch analyzes your data and the data of others similar to you to make recommendations of “movies you’ll love.” The Advertising Age article cites the Cinematch application as drawing millions of users to Netflix. I doubt this very much. I think Netflix was the right idea at the right time, and that alone garnered it millions of users. I personally think Cinematch is terrible at predicting movies I’ll love, and I prefer to base my picks on either Rotten Tomatoes or my favorite critic, Michael Phillips of the Chicago Tribune.
Anyway, that is rather beside my point. To learn more about the Cinematch application and algorithim, check out this article from Wired magazine. It explains how Cinematch works now (not so great) and why Netflix is offering $1 million dollars to the person or team that can build a better recommentaion engine.
(3 votes, average: 4.00 out of 5)
Loading ...
Subscribe to Comments
Follow us on Twitter
Get email updates
7 Comments, Comment or Ping
Renee Dupree
Ah, I like Netflix’s recommendation program. It’s not flawless, but it gets it right for me about 95% of the time. I’ve seen a lot of good movies that I probably wouldn’t have heard of otherwise.
Oct 29th, 2008
hrhmedia
Great post Jennifer, and definitely worth discussing tonight in class.
Oct 29th, 2008
jen.huss
I happen to agree with Jen. Netflix is terrible at recommending movies based on user ratings. They should model a project after Pandora and the Music Genome Project to analyze movies. Who wants to go after the $1 million with me?
Oct 29th, 2008
Brian Johnson
The Netflix recommendation engine is complex. I have actually read the forums where participants in the Netflix Challenge discuss the problems they run into in dealing with the data set and improving the recommendation engine. Nearly all of it is complex math talk, but it is interesting to read about how they are dealing with the data. Some teams have made significant improvements already.
I actually trust Netflix’s recommendations over critics. A critic is one person… I would rather get data from people who are statistically like me rather than somebody who may have been sick on the day they watched the movie.
I don’t dislike or like the recommendations… I only dislike or like the movies. If I liked every single movie I was recommended, I believe I would feel cheated because I should be seeing movies that aren’t always “perfect.” I should be pushed to discover new things and if everything is always 5 stars, then the movies are certainly not pushing me to the edge of my comfort zone.
Oct 30th, 2008
Mary Janisch
What’s interesting to me is the intense interest by the crowd in “cracking this nut.” According to the Leaderboard, there has been 31,747 valid submissions by 3,814 teams. So far Netflix has only paid out one progress prize of $50,000.
Oct 31st, 2008
crackerbelly
I also think it is interesting that the initial progress prize went to the guys from Bell Labs (alias: KorBell). They have now joined forces with the second place team from Commendo Research (alias: BigChaos). According to the leaderboard, they have pushed improvements in the algorithm to 9.44%. I am sure that this combined group of engineers is interested in the $1M prize money but I suspect even more important to them is the status and prestige that will come from winning. The resulting benefits from being able to put this on your resume will likely far exceed the prize money.
Nov 1st, 2008
adriana
I agree with Mary. Regardless of the outcome, the collaboration and ideation process of this competition is fascinating. It’s been running for such a long time! My favorite is the dad + daughter team of the ‘guy in a garage’ and how he’s been able to keep up with these math geeks for such a long time.
Nov 1st, 2008
Reply to “The Netflix Challenge”