A RBN-based recommender system architecture

With the widespread use of Internet, recommender systems are becoming increasingly adapted to resolve the problem of information overload and to deal with large amount of online information. Several approaches and techniques have been proposed to implement recommender systems. Most of them rely on flat data representation while most real world data are stored in relational databases. This paper proposes a new recommendation approach that explores the relational nature of the data in hand using relational Bayesian networks.