Mike's Blog

Netflix Recommender System

21 Aug, 2016
2 minute read

According to the article named “The Netflix Recommender System: Algorithms, Business Value, and Innovation.”, there are nine points in Netflix that can pick a perfect movie to users.

  1. Personalized Video Ranker (PVR): This is the most common recommend system, by users’ ranking. Users are able to rank a movie, and Netflix will recommend the movie to other users from high to low.

  2. Top-N video ranker. The top-N video ranker is a system can find best movies in each category and recommend to users. Some of them are based on PVR, but some of them are based on other places. By this system, users able to only focusing the head of the category, and it will improve the satisfies from users.

  3. Trending Now. The system can be easily understood from the name, the system will recommend a movie which is trending and also related to users habits.

  4. Continue watching. The continue watching ranker allow users keep watching TV show episodes or movies that users weren’t able to finish. Also by this ranker, Netflix can find what TV show and movies can attract a user to finish.

  5. Video- Video Similarity. This row will recommend videos similar with the videos users already watched. The system of this is not by users, all by a system from Netflix itself.

  6. Page Generation: Row selection and ranking. By this function, Netflix will recommend movies in a different type. Some accounts are shared by more than one users, and most of people like more than one genre movie.

  7. Evidence: in this row, users able to see the news about a movie that user just watched or ranked. This system will automatically find images and news for users.

  8. Search. 80% played videos are influenced by homepage, but there is still 20% played videos are based on research.

  9. Related work. In this function, the system can do machine learning techniques from the streaming data on the website to make the recommend system better.

By following the rules of the system, I think YouTube has the similar recommend system. But different to Netflix, YouTube is more focusing on the trending videos, so the system will recommend more trending videos than Top-N videos. And by type of videos watched on YouTube, the ads before videos will be changed as well.

Reference

  • Gomez-Uribe, C. A., & Hunt, N. (2016). The Netflix recommender system: Algorithms, business value, and innovation. ACM Transactions on Management Information Systems (TMIS), 6(4), 13.

  • (n.d.). Retrieved August 21, 2016, from http://www.imdb.com/title/tt0478970/?ref_=nv_sr_1