|Presentation Date||February 5, 2013|
|Topic(s)||You tube video recomendation algorithm|
YouTube wants it's users to watch more videos on their site. It gives them more profit in the form of add revenue and it brings in new videos which can then be monetized. This requires them to make a system that will recommend new videos for it's users to watch, and these recommendations must be good and quickly generatable.
My presentation looks at the factors YouTube has used to make these recommendations and how these changes have effected their user base. The old method relied on user input to define content and used these definitions to group similar content. This was found to be unreliable, people would game the system to gain more views. Their current method tracks user views and groups videos by which videos are watched in the same sessions.
YouTube FAQ: http://www.youtube.com/t/faq contains info on YouTube usage statistics
YouTube Creator blog: http://youtubecreator.blogspot.com/2012/03/changes-to-related-and-recommended.html information on changes and motivations for change to algorithm
Youtube partner info: http://support.google.com/youtube/bin/answer.py?hl=en&answer=2500737 Information for YouTube partners on how YouTube monetizes videos and calculates payouts.