{[ promptMessage ]}

Bookmark it

{[ promptMessage ]}

Netflix prize - rating results that the customer provides...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
Ben Olson Business Intelligence Netflix is currently holding a contest to see who can develop a system that beats their own. They use a program that tells the customer whether or not they will like a particular movie based on previously watched movies. This program runs using algorithms and formulas that give the customer vital feedback. For example, if I am really into fantasy/Action movies I will have a history of ordering fantasy/Action movies. The Netflix system will look up your order history and recommend to you a fantasy/Action film you haven’t seen and will probably enjoy. Netflix’s challenge is to see who can develop a system that can make better predictions than the Cinematch program currently being used by Netflix customers. Netflix provides the contestant with a lot of anonymous data that you then have to use to model your system. They also give you a prediction accuracy bar that has to be better than 10% of what Cinematch can do with the same data set. So this “accuracy bar” measures how close your systems’ results are to the actual
Background image of page 1
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: rating results that the customer provides. So if someone beats the 10% accuracy mark that Cinematch is currently running at, you get a serious money prize. The only catch is you have to disclose all or your information and source code to Netflix in order to receive the money. However, you could use this intellectual property to go and start your own business and possibly make a lot of money. For this reason Netflix offers a large amount of money in the amount of 1,000,000 dollars for the winner. It appears AT&T labs have developed two teams solely responsible for winning the contest. Bellkor and Korbell are their names, both of which are at the top of the leader board. The leader board ranks the teams by how much they have improved from their previous submit. Also, the amount of people involved in the contest seems to be enormous, I told the website to show me the top 500 teams and there were that many if not more....
View Full Document

{[ snackBarMessage ]}