Year of Graduation
Mechanism of Manipulation-Resistant Recommender System
Recommender system is an algorithm that uses preferences of its target user (expressed in form of ratings given to some items) and other users to predict how the target user would rate other items and recommend some of them. Some users may be interested in affecting performance of a recommender system by creating multiple artificial accounts in it. Information about social ties (trust) between users may be used for limiting influence of such attacks. The current work discusses the potential of application of trust in recommender systems and explores the properties of different trust metrics that allow collecting sufficiently large samples of user profiles which at the same time contain small number of artificial profiles created by the manipulator. Performance of algorithms that generate trust metrics has been modelled on random graphs and the trust network of Epinions recommender system.