Ontology-based user preferences Bayesian model for personalized recommendation

This paper proposes an ontology-based user preferences Bayesian model (UPOBM) for user preferences problem of traditional personalized recommendation. The model incorporates Bayesian network structure and knowledge of ontology to express the casual relations among contexts, user characteristics and user preferences. Taking a restaurant dishes recommendation under e-commerce as an example, the study adopts the method combining the probability reasoning of the proposed model with ontology rules to make recommendation. The experiment results show that the recommendation based on the proposed model is superior to the methods without Bayesian reasoning to user preferences in coverage and accuracy. © 2013 Binary Information Press.