Comparing the memory-based versus model-based recommenders
It is important to understand the strengths and weaknesses of both memory-based and model-based recommenders so that we can make the right choice according to the available data and the business requirements. As we saw in the previous chapter, we can classify recommender systems according to the data they are using and the algorithms that are employed.
First, we can talk about non-personalized versus personalized recommenders. Non-personalized recommenders do not take into account user preferences, but that doesn't make them less useful. They are successfully employed when the relevant data is missing, for example, for a user that is new to the system or just not logged in. Such recommendations can include the best apps of the week on the Apple App Store, trending movies on Netflix, songs of the day on Spotify, NY Times bestsellers, Billboard Top 10, and so on.
Moving on to personalized recommender systems, these can be further split...