← home

The Netflix Recommender System - Algorithms, Business Value, and Innovation

2022/01/18

  1. Personalize Video Ranker: PVR
    • Orders the entire catalog of videos (or subsets selected by genre or another filtering)
  2. Top-N Video Ranker
    • Find the best few personalized recommendations in the entire catalog for each member
  3. Trending Now
    • Short-term temporal trends, ranging from a few minutes to perhaps a few days, are potent predictors of videos that our members will watch.
    • 2 Types
      1. Those that repeat every several months (or yearly) yet have a short-term effect
        • Uptick of romantic video watching during Valentine's day
      2. One-off, short-term events, for example, a big hurricane with an impending arrival to some densely populated area
  4. Continue Watching
    • Sorts the subset of recently viewed titles based on the best estimate
      • Time elapsed since viewing
      • Point of abandonment (min-program, beginning or end)
      • Devices used
  5. Video-Video similarity
    • Unpersonalized algorithm that computes a ranked list of videos.
  6. Page Generation: Rows selection and ranking
    • Diverse selection of rows Personalized HomepagePersonalized Homepage

      Learning a Personalized Homepage - Netflix Tech Blog
      Recent Trends in Personalization: A Netflix Perspective
      Personalization at Netflix - Making Stories Travel
      Personalized Page Generation...
  7. Evidence
    • Evidence selection, select the most helpful evidence items for each member
    • For example,
      • Decide whether to show a movie that won an Oscar instead of a film is similar to another video recently watched by that member
      • Decide which image is the best support a given recommendation
  8. Search
    • 20% of hours streamed come from a search for videos
    • Users often search for videos, actors, or genres that are not in our catalog
    • Consist of a different algorithm, when the query is "fre"
      • Retrieve item (Frenemies)
      • Retrieve concept (French)
      • Retrieve relevant items given concept.

Reference

C. A. Gomez-Uribe and N. Hunt, β€œThe Netflix Recommender System: Algorithms, Business Value, and Innovation,” ACM Trans. Manage. Inf. Syst., vol. 6, no. 4, pp. 1–19, Jan. 2016, doi: 10.1145/2843948.

#netflix #recommender-system #recommender-system/videos