Introducing collaborative filtering
Collaborative filtering relies on mutual preferences, as it identifies items that a user might like based on how other similar users rated them. The central paradigm behind this approach is driven by the statement Show me the items people like me have chosen. I might find them interesting. There are two methods for implementing collaborative filtering systems: memory-based and model-based. In the first case, we utilize user rating data to compute the similarity between users or items. In the second case, models are developed incorporating machine learning (ML) algorithms to predict user ratings for unrated items. Let’s see both in more detail, starting with the memory-based approach.
Using memory-based collaborative recommenders
Before implementing the recommender, we need to sort out the data. One design choice is to utilize instances from reviewers who have made at least five evaluations. The reason is to exploit the most active users...