Netflix Prize Competition

Recently, the Netflix Prize, a three-year global analytics competition attracting teams from 186 countries, came to a photo-finish.  And when the dust settled, Opera, as part of a three-team collaboration called The Ensemble, stood at #1 on the Leaderboard and "first equals" in the contest overall.  While The Ensemble’s model performed as well as the ultimate Grand Prize winner, BellKor’s Pragmatic Chaos, BellKor’s model was submitted 20 minutes earlier, and, under contest rules, won.

The Ensemble was only one of two teams out of more than 41,000 entrants who succeeded in improving Netflix’s movie recommender program by over 10%, thereby qualifying for the contest’s grand prize. 

Opera entered the contest in order to test and hone techniques in new analytics. We made numerous advances that we are now applying in client assignments.  We developed a novel approach to K-Nearest Neighbors (“Learning KNN”) that significantly improved Netflix model performance and has subsequently proved valuable in real-world client applications.  In addition, our team refined and applied new ensemble techniques that significantly improved output (increasing by 2X any single model’s accuracy).

We are already successfully applying the lessons learned through our contest participation to once-intractable business problems. For example, our global Advanced Analytics team has created a powerful predictive engine to help a client in the food business recommend additional products to its customers, driving up sales by more than 15%.  For financial services firms, we are using these techniques in the risk, collections, and fraud prevention areas, outperforming traditional nonlinear models by substantial margins.  We are also employing advanced analytyics to value real estate portfolios and other investments, price products, and develop trading algorithms.

For the Netflix Prize, Opera drew on its global analytic talent, bringing together a team from our three scientific and technical Centers of Excellence in San Diego, Shanghai, and New Delhi.