Abstract
Personalized recommender systems provide relevant items to users from huge catalogue. Collaborative filtering (CF) and content-based (CB) filter- ing are the most widely used techniques in personalized recommender systems. CF uses only the user-rating data to make predictions, while CB filtering relies on semantic information of items for recommendation. In this paper we present a new approach taking into account the semantic information of items in a CF system. Many works have addressed this problem by proposing hybrid solu- tions. In this paper, we present another hybridization technique that predicts us- ers "preferences for items based on their inferred preferences for semantic in- formation. With this aim, we propose a new approach to build user semantic profile to model users‟ preferences for semantic information of items. Then, we use this model in a user-based CF algorithm to calculate the similarity between users. We apply our approach to real data, the MoviesLens dataset, and com- pare our results to standards user-based and item-based CF algorithms.