Hybrid Recommender System Based on Multi-Hierarchical Ontologies

Recommender Systems (RSs) are usually based in User Profiles (UP) to identify items of interest to a user, among the items of a usually vast collection. Traditional RSs are mostly based on ratings of items made by users and do not attempt to estimate the reasons that led the user to access these items. Furthermore, such systems may suffer from the lack of rating data, the so-called data sparsity. This paper proposes a hybrid recommender system that considers, besides the ratings of the users, a feature description analysis of the items accessed by the users. This analysis is based on ontological UP, described in accordance with a set of ontologies, one per feature. The use of ontologies provides a weak coupling between the proposed RS and the domain of the item to be recommended. The effectiveness of our proposal is demonstrated and evaluated in the movie domain using the MovieLens dataset. The experiments demonstrated an improvement in the quality of the recommendations and a greater tolerance to the data sparsity, compared to state-of-art systems.