A general rating recommended weight-aware model for recommendation system

Abstract

In recommendation system, the ratings represent the users’ preference and play an important role in recommending items to users. However, the ratings of items may be influenced by many factors, such as time (the latest ratings are more able to reflect the user’s current preferences), user familiarity (the more familiar a user to an item, the more reliable of rating he gives). So ratings should have different recommended weights in different circumstances. However, current recommendation algorithms ignore this problem and use the ratings indiscriminately, this affecting the accuracy of the recommendation system. In this paper, we proposed a general rating recommended weight-aware model, which can fuse all kinds of recommended weights naturally for item recommendation. We design a new rating weight-aware probability matrix factorization model, which can assign recommended weight to every rating to obtain precise recommendations. We conduct comprehensive experiments using the real-world datasets. Experimental results show that the rating-aware recommendation model outperforms state-of-the-art latent factor models with a significant margin.