Robust Recommendation Algorithm Based on the Identification of Suspicious Users and Matrix Factorization
Huawei Yi · Journal of Information and Computational Science · 2014
The existing recommendation algorithms have lower robustness against shilling attacks. With this in mind, in this paper we propose a robust recommendation algorithm based on the identification of suspicious users and matrix factorization. We first give the computational methods of Deviation Degree about the Number of Ratings (DDNR) and Average Similarity of Neighbors (ASN) according to the distribution of users’ ratings. Then we identify suspicious users based on the differences of computational results of users’ DDNR and ASN. Finally, we introduce the matrix factorization technology which incorporates indicator function and the identification results of suspicious users to make recommendations for users. Experimental results show that the proposed algorithm not only improves the recommendation accuracy, but also has better robustness.