A Balanced Collaborative Filtering Approach Incorporating with Conformity

Abstract

Collaborative filtering can estimate users' ratings for unvisited items based on the opinions about items implied in their observed ratings. The issue of sparsity induced by the insufficiency of rating is a key factor impacting the recommendation accuracy. Aiming at the issue of sparsity, a balanced collaborative filtering approach is proposed in this work. According to the conformity of users, the proposed approach employs the target item's general rating and personalized rating to predict the rating for it, with adjusting importance of both types of rating.