Collaborative filtering in a recommender system is the act of generating personalised recommendations through the data of similar users. This data is for example the interactions of a users that is inline with the interactions of the user we want to recommend too.
In collaborative filtering user to user recommendations and item to item are most common. This mean that we recommend similar users for a user or similar items to another item. One could have Item to user recommendations where items are recommended to users.
In essence the following recommendations are usually made with collaborative filtering: Others also purchased, Similar users watched, Other people liked.
Challenges
Cold starts, when there is not a lot of data could pose a problem for this approach. Semantic recommendations might be a great way to alleviate some pains.
Data
A lot of times the data of the interactions goes through some data anonymization process so no Personal Identifiable information (PII) is saved or used.
Sources
Recommender system can be either personalized or non-personalized. Non-personalized system can be simpler but personalized system tends to work better as it caters to the needs of each individual user. Collaborative filtering is a common method of personalized recommender system which filters information such as interactions data from other similar users. Since it works by predicting user ratings, it is considered as performing regression task. There are two general types of collaborative filtering:
— ^2404ed from Collaborative Filtering in Recommender System: An Overview
User to user collaborative filtering basically operates under the assumption that users who gave similar ratings to a certain item are likely to have the same preference for other items as well. Therefore this method mainly relies on finding similarity between users. However, in some cases, user preference might be to abstract to be broken down. This is where item to item collaborative filtering comes in handy. Here, similarity between items is used instead of similarity between users. In this article, we’ll be focusing on user to user collaborative filtering.
— ^262887 from Collaborative Filtering in Recommender System: An Overview
First type of the models Recombee adds to its model ensembles are Collaborative Filtering models, which are built from collected user-item interactions, such as detail-views or purchases. By analyzing behavioral patterns across the whole userbase, the recommendations are based on extracting interactions similarities between users, items, or both. “Similar users also liked” or “others also purchased” are both examples of CF-based recommendations. Recombee uses following CF models: Matrix Factorization, Nearest Neighbor methods, and Association Rules.
— ^3e2cf6 from What Makes Our Recommendation Technology Unique