The ranking problem

Ranking is the more intuitive formulation of the recommendation problem. Given a set of n items, the ranking problem tries to discern the top k items to recommend to a particular user, utilizing all of the information at its disposal.

Imagine you are Airbnb, much like the preceding example. Your user has input the specific things they are looking for in their host and the space (such as their location, and budget). You want to display the top 10 results that satisfy those aforementioned conditions. This would be an example of the ranking problem.

It is easy to see that the prediction problem often boils down to the ranking problem. If we are able to predict missing values, we can extract the top values and display them as our results.

In this book, we will look at both formulations and build systems that effectively solve them.