Affiliations: [a] School of Information Science and Engineering, Yanshan University, Qinhuangdao, China
| [b] School of Science, Yanshan University, Qinhuangdao, China
Correspondence:
[*]
Corresponding author. Fuzhi Zhang, School of Information Science and Engineering, Yanshan University, Qinhuangdao, China. Tel.: +86 335 8057078; Fax: +86 335 8074806; E-mail: [email protected].
Abstract: Collaborative filtering recommendation algorithms based on traditional matrix factorization model have poor robustness and low recommendation accuracy when facing shilling attacks. To address this issue, we propose a novel robust recommendation method based on kernel matrix factorization. We first construct a robust kernel matrix factorization model for collaborative recommendation by using kernel mapping of the rating matrix and kernel distance, and regulate residual error with the scale factor, which can enhance the power of the model’s anti-attack and realize the robust estimation of user feature matrix and item feature matrix. Then we introduce kernel distance to compute the similarity between users in order to improve the credibility of user similarity and reduce the influence of attack profiles on the recommendation results. Finally, we devise a robust collaborative recommendation algorithm based on the kernel matrix factorization model. Experimental results show that our algorithm can improve the robustness and accuracy compared with the existing algorithms.
