Analytics Stories

CHAPTER 29How Does Netflix Recommend Movies and TV Shows?

We have so many choices in today's world. What book should we read next? What movie should we rent? What hot new song should we download to our smart phone? Collaborative filtering is the buzzword for methods used to “filter” choices using the collective intelligence of other people's product choices. The web has made it easy to store the purchasing history and preferences of consumers. The question is how to use this data to recommend products to you that you will like but didn't know you wanted. If you ever streamed a movie from a Netflix recommendation, bought a book from an Amazon.com recommendation, or downloaded a song from iTunes from a GENIUS recommendation, you have utilized a result generated by a collaborative filtering algorithm.

In this chapter, we use simple examples to illustrate the key concepts used in collaborative filtering. We will discuss user-based and item-based collaborative filtering algorithms. To illustrate the difference between these two methods, suppose we have not seen the movie Bohemian Rhapsody and want to know if we would like it. In user-based collaborative filtering, we look for moviegoers whose rating of movies we have seen is most like ours. After giving a heavier weighting to the most similar moviegoers who have seen Bohemian Rhapsody, we can use their ratings to generate an estimate of how well we would like Bohemian Rhapsody. In item-based collaborative filtering (first used by ...