Chapter 4. Ratings and how to calculate them
- Creating user-item matrices.
- Revisiting explicit ratings to discover why they aren’t always good.
- Diving into the mystery of implicit ratings and its creation.
- Learning about an implicit ratings function that translates evidence into ratings.
In this chapter, you’ll transform your users’ behavior to a format that you can use as input for the recommender algorithms. You’ll start by looking at the user-item matrix, which is where most recommender algorithms start. Then you’ll take another look at explicit ratings, the ratings that users add themselves. Implicit ratings are the core of your system, and you’ll look at those next: first, you’ll review what they are and then you’ll learn how to calculate them from your evidence.
Figure 4.1 shows the flow of data we’ve talked about so far. Data collection happens when visitors interact with the site. Preprocessing is what you’re going to do in this chapter. Model building and recommendation construction are handled in later chapters.
In this chapter, you’ll convert web behavior to content ratings, which you’ll use for the recommenders in later chapters. This type of rating is called implicit because they’re deduced.