Collaborative filtering is the predictive process behind recommendation engines. Recommendation engines analyze information about users with similar tastes to assess the probability that a target individual will enjoy something, such as a video, a book or a product. Collaborative filtering is also known as social filtering.
Collaborative filtering uses algorithms to filter data from user reviews to make personalized recommendations for users with similar preferences. Collaborative filtering is also used to select content and advertising for individuals on social media.
Three types of collaborative filtering commonly used in recommendation systems are neighbor-based, item-to-item and classification- based.
In neighbor-based filtering, users are selected for their similarity to the active user. This similarity is determined by matching users who have posted similar reviews. Based on the previous similarity, it is presumed that future likes and dislikes will also be similar. From the average rating of the group, recommendations are made for the active user.
An item-to-item filtering process uses a matrix to determine the likeness of pairs of items. Item-to-item processes then compare the current user’s preference to the items in the matrix for similarities upon which to base recommendations.
A classification-based collaborative filtering system recommends things based on how similar users liked that classification or genre. It is assumed that users that enjoy or dislike similar experiences within a classification will also enjoy others within that classification.
Some collaborative filtering systems are memory-based, like neighboring- and item-to-item models, which compare similarities of users or items. Others are model-based, using machine learning to compare dissimilar items. Model-based systems may use algorithms such as the Markov decision process to predict ratings for items that have not yet been reviewed. Hybrid systems include features of both memory-based and model-based filtering.
Recommendation systems are used to provide suggestions for all kinds of websites and services. Still, they can encounter a number of difficulties. The sparsity of ratings is one of the main hurdles to collaborative filtering’s usefulness in systems with many items. New items also tend to be difficult to provide recommendations for. Under new recommendation systems, it is hard to provide good recommendations before enough users have entered reviews. At the same time, however, too many user ratings can be challenging to some systems because they make for huge data sets.