An item-based collaborative filtering approach based on balanced rating prediction

As a widespread approach in recommender systems, item-based collaborative filtering can predict an active user´s interest for a target item based on his interest and the ratings for those similar items to his visited items. As the effect of human´s conformity psychology, an individual user´s judgment usually tends to follow the general view. The majority of existing item-based collaborative filtering approaches emphasizes the personalized factor of recommendation separately, but ignores the user´s general opinions about items. Aiming at this issue of unbalanced recommendation, this paper proposes a refined item-based collaborative filtering approach which employs a balanced rating prediction method incorporating an individual´s personalized need with the general opinions. The experimental result shows an improvement in accuracy in contrast to the classic item-based collaborative filtering.