Pairwise Factorization Machines for Personalized Ranking with Implicit Feedback

Abstract

Pairwise learning is a novel technique for personalized recommendation with implicit feedback. Pairwise learning algorithms can well learn user's preference, from not only the observed user feedbacks but also the underlying interactions between users and items. It also has long proved that incorporate contextual information in model can further improve the accuracy of recommendation. In terms of the problem of personalized recommendation for implicit feedback and how to incorporate users' contextual information in recommendation, this paper proposes a recommendation model combined with pairwise learning and factorization machine. First of all, we use the method of pairwise learning to solve the problem of negative feedback missing under implicit feedback scenario, and then choose the factorization machine as ranking function to model the user's contextual information, and provide personalized recommendations for different users according to the model score. Experiments also show that the model proposed in this paper is better than the other three contrast algorithms in terms of ranking metrics such as MAP and NDCG.