A Trust Prediction Method for Recommendation System

With the increasingly fierce competition of ecommerce platform, the trust relationship among users has become the research hotspot of recommendation system. In this paper, trust relationship is introduced into the recommendation system, and a multi-source attribute trust prediction method based on improved D-S evidence theory is proposed. Firstly, the user attributes are analyzed quantitatively according to the user individual data, and four attributes are selected. Then, the qualitative attributes is obtained by discretizing the quantitative attributes. At last, the attributes evidence is fused repeatedly using the weight assignation method to get the trust relationship strength triple. In the simulation, the sufficiency of attribute evidence and the trust prediction result are verified by the sevenfold cross-validation method. In addition, the proposed method is compared with other methods of machine learning, and the result proves the superiority of the proposed trust fusion mode.