The first Netflix award succeeded in solving a huge challenge, accurately predicting tastes for viewers who provided more than 50 ratings. The next million-dollar prize aims to recommend movies for customers who don't do movie ratings often or at all, requiring predictions to be made using some of the geographic and behavioral data that hides viewers' tastes. Again, the winner will need to make their algorithm public. If this problem can be solved, Netflix would be able to start recommending movies to new customers very quickly, without waiting for customers to provide a lot of ratings data before making recommendations.
The new dataset for the contest has 100 million pieces of data, including ratings data, customer age, gender, zip code of area of residence, and previously viewed movies. All of the data is anonymized and there is no way to associate it with any of netflix's customers.
Unlike the first Grand Prix, there was no set goal for the competition. 500,000 dollars was awarded first to the team that took the lead in six months, and another 500,000 dollars was awarded to the team that took the lead after 18 months.
The recommendation engine is a key service at Netflix, where more than 10 million customers are able to rate movies on a scale of 1-5 on a personalized web page.Netflix puts those ratings in a huge dataset of more than 3 billion entries.Netflix uses recommendation algorithms and software to identify movies that are likely to be viewed by people with similar tastes in the ratings that viewers are likely to give a movie. For two years, Netflix has been using contestants to improve the efficiency of its movie recommendations, which has been well-received by many movie critics and subscribers.
Dr. Rick Hangartner, chief scientist of Strands' recommendation engine, writes: "In the short term, search engines will increasingly incorporate simple recommendation techniques to deal with close query terms (e.g., "What you're looking for is this, and based on similar queries/other people's searches, what you're probably looking for is this.") But in the long run, and more than the search industry and search technology, recommendation technology will be even more pervasive."