One of the freelancing hats I wear these days is graphic design for the California Institute of Technology’s award-winning Engineering & Science magazine, and its latest issue contains a really fascinating article on the Netflix Prize contest (2006-’09) that awarded a million dollars to the person/team who best improved the company’s algorithm for predicting its user ratings.
I’m sure most readers here have received their fair share of movie predictions from any number of websites, ranging from the accurate to the absurd. A few months ago, Amazon.com actually sent me this email: “As someone who has purchased or rated The Philadelphia Story, you might like to know that Furry Hamsters From Hell is now available.” This wasn’t a practical joke, it was a real attempt to persuade me to click on their website and spend $19.95. On the other hand it sometimes gets it right, like when it told me that based on my previous purchases, I might be interested in the upcoming Alamar (2009) from Film Movement.
“Recommend a Movie, Win a Million Bucks” (it’s a PDF) is written by Joseph Sill, an analytics consultant who spent “the better part of a year” competing with programmers around the world, hoping to discover the right statistical combination that would generate the most accurate predictions by July 26, 2009. The article is a fun–even suspenseful–and informative read, a crash course in machine learning rife with movie references.