Low-rank and sparse matrix factorization with prior relations for recommender systems
- PDF / 1,780,057 Bytes
- 15 Pages / 595.224 x 790.955 pts Page_size
- 56 Downloads / 183 Views
Low-rank and sparse matrix factorization with prior relations for recommender systems Jie Wang1 · Li Zhu1 · Tao Dai1 · Qiannan Xu1 · Tianyu Gao1 Accepted: 13 October 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract The explosive growth of data has caused users to spend considerable time and effort finding the items they need. Various recommender systems have been created to provide convenience for users. This paper proposes a low-rank and sparse matrix factorization with prior relations (LSMF-PR) recommendation model, which predicts users’ ratings for items through a sum of the learned low-rank matrix and sparse matrix. Thus, unlike traditional matrix factorization approaches, our method can alleviate the error propagation produced by intermediate outputs. The LSMF-PR integrates user relationships and item relationships as prior information. User relationships in different recommendation scenarios are extracted by the corresponding social relations of the users, and item relationships are obtained from the similarity of the item content. Therefore, the sparsity and cold start problems can be effectively reduced with prior information. Furthermore, our model has better interpretability since it reveals the low-rank and sparse features of the ratings. Experiments are conducted on four real-world datasets to validate the performance of our proposed method. Keywords Recommender system · Low-rank and sparse matrix factorization · Prior information
1 Introduction The exponential development of Internet has created information overload problems for people. Large databases lead to wastes in energy and time when users find appropriate items. To alleviate this problem, a recommender system (RS) has been developed as a key solution in various fields, such as e-commerce [1] and web service [2]. An advanced recommender system could provide suggestions that meet the needs of users. Collaborative filtering (CF) is a typical Li Zhu
[email protected] Jie Wang wangjie [email protected] Tao Dai [email protected] Qiannan Xu [email protected] Tianyu Gao [email protected] 1
Xi’an Jiaotong University, Xi’an, China
recommendation method that recommends items for users by using a rating matrix. The effectiveness of CF relies on the assumption that users with common interests will likely give similar ratings on an item. The user-item (U-I) relations are preserved in a U-I matrix. This kind of relation has been widely exploited in various scenarios [3, 4] due to its efficiency and simplicity. One classic method of CF is matrix factorization [5, 6], which factorizes the original rating matrix into two low-rank matrices. One matrix indicates the interest of users. The other denotes the hidden features of items. The users ratings for items are then predicted by the products of the user latent factor vectors and item latent factor vectors. However, the lack of interactive information among users and items makes this method suffer from sparsity and cold start problems. An effective way to solve t
Data Loading...
