How Netflix’s Recommended Movies are (Page)Ranked
Similar to Google’s PageRank algorithm used to determine the order of a query’s search results, Netflix also has an algorithm that utilizes a method known as collaborative filtering to determine each users’ recommend movies. The Coursera lesson referenced below from a Princeton University entitled “Networks: Friends, Money, and Bytes” provides insightful commentary on the way Netflix organizes its recommended movie list for users.
Collaborative filtering takes into account what users like as well as what moveis users with similar interests like, and makes a decision based on that. The question to consider is what constitutes “similar” interests. Because many of the top companies’ complete algorithms are not public knowledge, we can only examine the inputs and outputs of Netflix’s engine. In terms of inputs, Netflix examines the user ID and movie ID, indexes both of them, and uses those indexes to retrieve a specific rating of the movie. It also considers the specific time that the search of the movie was initiated by the user. The text description of the movie and how it relates to the users’ previously viewed movies also plays an important role in the algorithm. The output of the algorithm involves a predicted rating, which is how much the system believes the user will enjoy that specific movie.
Netflix then uses uses each movie’s predicted rating to determine the possible customer satisfaction and predictive effectiveness and order the recommended movies for each user. The customer satisfaction, though hard to quantify, relates to if the user enjoyed the movie they watched. The prediction effectiveness, on the other hand, indicates whether the user even watched a recommended movie or not. Netflix’s ranking system is based on mathematical calculations just like Google’s PageRank algorithms. They both exhibit methodical approaches to the problem and keep the user in mind when outputting results. It is interesting to see how these designs are broken down and the integral roles algorithms like play in users’ everyday lives.
Resource: Coursera