A collaborative and semantic-based approach for recommender systems

The constant growth of the Internet has made recommender systems very useful to guide users coping with a large amount of data. In this paper, we present a domain independent collaborative and semantic-based recommender system which uses distinct and complementary modules. The approach targets users with various interests and is based on: (i) a collaborative module using association rules in order to mine a set of rules for the target user, (ii) a semantic module using the domain ontology to reason about items, (iii) a frequency module using the frequency of the item features in order to discover additional items to be recommended. Unlike numerical approaches, applying these different modules separately provides a multi-view basis to explain the recommendations proposed to the user. Our recommendation system has been tried out on the MovieLens dataset and the results that have been evaluated by a set of volunteers are promising.