A Survey of Collaborative Filtering Techniques

Abstract

As one of the most successful approaches to building recommender systems, collaborative filtering (CF) uses the known preferences of a group of users to make recommendations or predictions of the unknown preferences for other users. In this paper, we first introduce CF tasks and their main challenges, such as data sparsity, scalability, synonymy, gray sheep, shilling attacks, privacy protection, etc., and their possible solutions. We then present three main categories of CF techniques: memory-based, model-based, and hybrid CF algorithms (that combine CF with other recommendation techniques), with examples for representative algorithms of each category, and analysis of their predictive performance and their ability to address the challenges. From basic techniques to the state-of-the-art, we attempt to present a comprehensive survey for CF techniques, which can be served as a roadmap for research and practice in this area.

A Novel Embedding Model for Relation Prediction in Recommendation Systems

“…The recommendation algorithms can be divided into three classes: content-based [26], collaborative filtering (CF) [22], [30] and hybrid recommendation algorithms [1]. The content-based approaches analyze the content feature of the items and the attributes of the users, and then measure similarity between the users and items for recommendation, regardless of the users's rating/selection history.…”

“…Consequently, they suffer from the sparsity problem if rating/selection history is short. CF algorithms can be further divided into two categories: memory-based CF [23], Item-KNN [28] and model-based CF [30]. The hybrid approaches combine the content-based and CF methods to avoid their limitations.…”

Semantically-enhanced advertisement recommender systems in social networks

“…In this way, the things with the most astounding anticipated appraisals are prescribed. Concerning the comparability of clients, [262] and [263] state that for the most part either a relationship based likeness (e.g., Pearson connection coefficient) or a cosinesimilitude measure in view of client profile vectors is connected. With respect to thing based sifting, the objective is to locate the most comparable things taking into account the client profiles of the clients who appraised these things [264].…”

Big Data, Privacy, and Trusted Web: What Needs to Be Done

“…This results are noteworthy, because traditional privacy attacks were based on aggregating information from multiple datasets. Such methods were based on collaborative filtering [46] and enabled an efficient and highly reliable characterization of a person from a few data. The underlying technology is quickly advancing [47], and it may give service providers, such as mobile phone, Internet television, or social gaming centers an unprecedented amount of personal information.…”