Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
5103441 | Physica A: Statistical Mechanics and its Applications | 2017 | 9 Pages |
Abstract
As the explosion growth of Internet economy, recommender system has become an important technology to solve the problem of information overload. However, recommenders are not one-size-fits-all, different recommenders have different virtues, making them be suitable for different users. In this paper, we propose a novel personalized recommender based on user preferences, which allows multiple recommenders to exist in E-commerce system simultaneously. We find that output of a recommender to each user is quite different when using different recommenders, the recommendation accuracy can be significantly improved if each user is assigned with his/her optimal personalized recommender. Furthermore, different from previous works focusing on short-term effects on recommender, we also evaluate the long-term effect of the proposed method by modeling the evolution of mutual feedback between user and online system. Finally, compared with single recommender running on the online system, the proposed method can improve the accuracy of recommendation significantly and get better trade-offs between short- and long-term performances of recommendation.
Keywords
Related Topics
Physical Sciences and Engineering
Mathematics
Mathematical Physics
Authors
Xiaoyu Shi, Ming-Sheng Shang, Xin Luo, Abbas Khushnood, Jian Li,