AbstractsComputer Science

In today’s consumer driven world, people are faced with the problem of plenty. Choices abound everywhere, be it in movies, books or music. Recommender systems spare the user the frustration of searching for the proverbial needle in the haystack by offering recommendations based on a user’s personal preferences. In this thesis, a generic framework and algorithms for building concept-based recommender systems are presented. A concept-based approach leverages the deep structure of a rating-database and reveals complex, higher level inter-relationships between entities in the data. The algorithms encode user preferences from a ratings database into concepts using collaborative filtering, organize concepts into lattices efficiently, and enable fast querying of the lattices for recommendations. We apply our algorithms on two real-world datasets and demonstrate their capabilities in generating quality recommendations in real-time.