There’s a serious amount of money to anyone who can prove that the answer is yes, and rather surprisingly it is online DVD rental service Netflix that is posing the question and providing the cash.
It all revolves around the movie recommendation system that Netflix developed, CinematchSM, that can predict whether a user will like a film based upon how much they have liked or disliked previously rented ones. The concept is nothing new, Amazon have such a personal recommendation system and so do pretty much all the players in the online DVD rental business.
The difference is that nobody else is offering a million bucks, yes you read that right, for the person or persons who can best increase the predication accuracy of the existing system, by a factor of at least 10%. Measuring this accuracy against the same set of training data, a 2Gb dataset containing around 17,000 movie titles and 100,000,000 (personal data cleansed) user ratings, the metric is simply how close predicted ratings of movies match subsequent actual ratings.
The catch, and there has to be one I guess, would appear to be that the prize is only handed out if you share your methodology with (and non-exclusively license it to) Netflix and the world. But hey, for a million dollars I suspect that won’t put too many folk off.
If code optimization is your thing, and you live and breathe algorithms, you may want to give it a shot. If you do then you will be joining the 11832 current contestants from 9627 teams spread across 102 different countries. Better yet, why not start a thread in the Software Developer’s Lounge and get a DaniWeb team together? It may seem as if there is no great rush to enter, seeing that the competition will run to ‘at least’ October 2, 2011, but it has only just started (October 2) and there are yearly progress prizes of $50,000 to the team which shows the most improvement over the previous year’s accuracy figures.
It’s free to enter, could just help you hone your own skills during the process, and is one heck of a fun way to get into some serious machine learning and recommendation systems education.