A Review of Trust-Aware Recommender Systems Based on Graph Theory

The Web is currently characterised by user contribution. As a result, content is generated in an uncontrolled way leading to the so-called “information overload”. The role of information filtering techniques and recommender systems is to give a solution to this problem by taking into account user preferences, and/or other context of information to present the content. Furthermore, word-of-mouth and trust plays a key role in the decision-making process of a person, while ratings, comments, opinions and tags are increasingly gaining popularity in social networks. The aim of this paper is to study the current approaches of trustaware recommender systems with a focus on graph based models and to identify possible gaps for providing insights and directions for future researches. In the beginning there is an introduction with presentation of traditional recommendation approaches along with their limitations. Then there is a review of trust definitions and properties. It follows a review of current trust-aware approaches which are classified in five major categories according the technique they use. In the last section there is a focus on graph models of trust-aware recommender systems which are compared according four categorisation criteria. Finally the study identifies the gaps in the current literature of graph based models and proposes areas for future research.