Enhancing collaborative filtering using semantic relations in data

Abstract

Recommender Systems (RS) pre-select and filter information according to the needs and preferences of the user. Users express their interest in items by giving their opinion (explicit data) and navigating through the webpages (implicit data). In order to personalize users experience, recommender systems exploit this data by offering the items that the user could be more interested in. However, most of the RS do not deal with domain independency and scalability. In this paper, we propose a scalable and reliable recommender system based on semantic data and Matrix Factorization. The former increases the recommendations quality and domain independency. The latter offers scalability by distributing treatments over several machines. Consequently, our proposition offers quality in user’s personalization in interchangeable item’s environments, but also alleviates the system by balancing load among distributed machines.