Encoding for recommender systems

So far, in the previous sections, we have encoded an image. In this section, we will encode users and movies in a movie-related dataset. The reason for this is that there could be millions of users as customers and thousands of movies in a catalog. Thus, we are not in a position to one-hot encode such data straight away. Encoding comes in handy in such a scenario. One of the most popular techniques that's used in encoding for recommender systems is matrix factorization. In the next section, we'll understand how it works and generate embeddings for users and movies.

Getting ready

The thinking behind encoding users and movies is as follows:

If two users are similar in terms of liking...