Recommendation Engine

Imagine you have three movies**:

  1. Army Army
  2. Love’s Heartstrings
  3. Road Racer

** Not intended to be real movies.

Imagine that you have a list of three people: 

  1. Joe gives a rating of 5 out of 5 to Army Army and Road Racer. He gives Love’s Heartstrings a 1 out of 5.
  2. Sam gives a 4 out of 5 to Love’s Heartstrings and Road Racer, but a 2 out of 5 to Army Army. 

Here’s a table of their results. 

What movie do you think that Sam and Joe will recommend to Elliot? Most likely, Joe and Sam will recommend that Elliot see Road Racer. Both Joe and Sam liked it. Their opinions intersect on that move.

A Simple Example 

Above is a simple example of a recommendation engine. Each person gives input into the movie that they have watched. Based on the intersection of the movies that people have liked, a recommendation is made for the next person. 

Recommendation engines get better when there is more information. In our example, we don’t know whether Elliot likes war movies better, or love stories. If we knew this, the recommendation might change. Also, the number of people recommending is very small: this means that it is hard to get a good understanding of what a large group of people think.