Researches And Application On Trust Based Recommendation With User Preference

Trust-based recommendation has been an important research topic in recommender system research. In this thesis, in order to solve the problem of user-tag preference sparsity and that it is hard to recommend when the number of trusted users is not enough, we focus on a trust-based recommendation model with user-tag preference. Thus, we can recommend users’ interesting resources to them.The main researches in this thesis include:1) Researches on related trust models and proposal of a new trust model called MPDTrust, considering both implicit direct trust and indirect trust. Implicit direct trust in MPDTrust is based on rating information of users considering the number of users’ co-rated items, rating time and rating error. Indirect trust in MPDTrust is based on TidalTrust considering the maximum trust propagation distance.2) Researches on user-tag preference models. Not only users’ratings and tags were considered in calculating user-tag preferences, but also time information was considered. In addition, we present a method of expanding user-tag preferences with considering the correlation of tags and tag preferences of trusted users.3) Proposal of a trust-based recommendation model with user-tag preference, called TUPTR.4) Design and implementation the TTUPFilm recommender system.Our main contributions in this thesis are:1) Proposal of a trust model based on maximum trust propagation distance, which considers implicit direct trust.2) Proposal of a method of extending user-tag preferences considering the correlation of tags and tag preferences of trusted users.3) Proposal of a trust-based recommendation model with user-tag preference, called TUPTR.The research in this thesis would be valuable to the studies of recommendation based on trust network and user preference.