Constructing a user similarity network to remove adverse influence of popular objects for personalized recommendation

摘要

Nowadays, personalized recommender systems have become more and more indispensable in a wide variety of commercial applications due to the vast amount of overloaded information accompanying the explosive growth of the internet. Based on the assumption that users sharing similar preferences in history would also have similar interests in the future, user-based collaborative filtering algorithms have demonstrated remarkable successes and become one of the most dominant branches in the study of personalized recommendation. However, the presence of popular objects that meet the general interest of a broad spectrum of audience may introduce weak relationships between users and adversely influence the correct ranking of candidate objects. Besides, recent studies have also shown that gains of the accuracy in a recommendation may be frequently accompanied by losses of the diversity, making the selection of a reasonable tradeoff between the accuracy and the diversity not obvious. With these understandings, we propose in this paper a network-based collaborative filtering approach to overcome the adverse influence of popular objects while achieving a reasonable balance between the accuracy and the diversity. Our method starts with the construction of a user similarity network from historical data by using a nearest neighbor approach. Based on this network, we calculate discriminant scores for candidate objects and further sort the objects in non-ascending order to obtain the final ranking list. We validate the proposed approach by performing large-scale random sub-sampling experiments on two widely used data sets (MovieLens and Netflix), and we evaluate our method using two accuracy criteria and two diversity measures. Results show that our approach significantly outperforms the ordinary user-based collaborative filtering method by not only enhancing the recommendation accuracy but also improving the recommendation diversity.