Context Aware Recommender System Algorithms: State of the Art and Focus on Factorization Based Methods

• Main Authors: Fatima Zahra Lahlou, Houda Benbrahim, Ismail Kassou
• Format: Article
• Language: English
• Published: Ecole Mohammadia d'Ingénieurs 2017-11-01
• Series: Electronic Journal of Information Technology
• Subjects: Context Aware Recommender Systems; Matrix Factorization; Tensor Factorization; Factorization Machines; Machine Learning
• Online Access: http://www.revue-eti.net/index.php/eti/article/view/116
• ISSN: 1114-8802; 1114-8802

Context Aware Recommender Systems (CARS) have become an important research area since its introduction in 2001 by (Herlocker and Konstan, 2001) and (Adomavicius and Tuzhilin, 2001). According to the classification of Adomavicius et al. (Adomavicius and Tuzhilin, 2011), there are three main categories of CARS algorithms: pre-filtering, post-filtering, and contextual modeling ones. Surprisingly, until the year of 2010, almost no CARS modeling algorithms were proposed, even though contextual modeling recommender systems can theoretically accept more dimensions as contextual variables (Karatzoglou et al., 2010). Starting from 2010, many contextual modeling CARS algorithms were proposed, most of them are built on factorization models. In this paper, we first present a state of the art of domain independent CARS algorithms listed following a chronological order. Then we study factorization models used for the Context Aware Recommendation task and suggest some possible research directions for developing more performing contextual modeling CARS algorithms.