基于张量分解的异构隐式反馈推荐算法研究

Most of implicit feedback based recommendation algorithms only utilize little auxiliary feedback, which leads to the inaccuracy recommendation in data sparsity. However, several user actions on e-commerce, such as click, wanted, purchased, can provide extra potential and valuable information for recommender systems. To make full use of these actions, this paper proposes a tensor-based recommendation algorithm using heterogeneous implicit feedback. This scheme exposes the hidden dependency among users, items and actions and breaks the limitation of user-item matrix. Moreover, it also considers the social information as regularization term to obtain trust relationship between users and their friends. The experimental results on a real dataset show our proposed algorithm outperforms other compared methods, and improving the performance of recommender system effectively