Yesterday I wrote about the Netflix prize — $1,000,000 awarded to the team BellKor’s Pragmatic Chaos for creating an algorithm that was 10.06% better at recommending movies that customers would like.
Improvements came quickly and then bogged down. Here are the highlights:
- 2007 $50,000 progress prize — BellKor with an 8.43% improvement
- 2008 $50,000 progress prize — BellKor in BigChaos with a 9.44% improvement. This team was a combination of the two front runners BellKor and Big Chaos
- 2009 $1,000,000 grand prize — BellKor’s Pragmatic Chaos at 10.06%. This team was a combination of BelKor in Big Chaos and Pragmatic Theory. Another collaborative team, the Ensemble (a merger of the Grand Prize Team and Opera Solutions and Valdelay United), tied but their final submission was submitted 20 minutes later.
“This has been one of the wonderful discoveries in the competition, that blending teams can lead to substantial gains…” said Chris Volinsky, a scientist at AT&T Research and a member of BellKor’s Pragmatic Chaos Team. Blending different technical skills (statistical and machine-learning techniques) “only works well if you combine models that approach the problem differently. That’s why collaboration has been so effective, because different people approach problems differently.”
Some of the factors that affected predictions were:
- people rate movies they saw a long time ago differently than the ones they saw recently
- movie watchers tend to rate movies differently on Fridays versus Mondays
- a rating given on a Monday is a poor indicator of other movies the viewer will like
Other companies are also using crowdsourcing to solve real problems. Check out these websites to see some of the opportunities offered through these clearinghouse sites:
What problem would you like to solve by offering a prize?