Modeling Recommender Systems from Preference and Set-Oriented Perspectives

Recommender systems aim to recommender right products to right customers. Considering the similarity between recommender systems and information retrieval, researchers have extended results of information retrieval to recommender systems. These motivate us to further discuss the recommender systems by using results of information retrieval. This paper borrows the results of information retrieval to discuss the traditional recommender system methods by using the preference relation defined on the cartesian product of a set of customers and a set of products. Instead of providing a new recommender system algorithm, this paper aims to provide a framework to interpret recommender systems from a set-oriented view and wish to stimulate new recommender algorithms.