Personalized collaborative filtering: a neighborhood model based on contextual constraints

Abstract

In this paper, we propose a recommender system approach which considers contextual information from users and items in order to improve the accuracy of a neighborhood-based collaborative filtering algorithm. One advantage of our model is the possibility to bias the users' similarity computation according to a contextual constraint, such as the group of individuals who share the same demographic information, or the set of users with whom the user is interacting at the moment. The proposal represents the first steps towards the development of a group recommender system model. We provide an evaluation of our method with the MovieLens dataset, and compare our approach against other known techniques reported in the literature.