Recommendation algorithm of probabilistic matrix factorization based on directed trust. (July 2021)

• Record Type: Journal Article
• Authors: Xu, Shangshang; Zhuang, Haiyan; Sun, Fuzhen; Wang, Shaoqing; Wu, Tianhui; Dong, Jiawei
• Is Part Of: Computers & electrical engineering. Volume 93(2021)
• Publication Date: 2021-07
• Subjects: Recommender system -- Social recommendation -- Probabilistic matrix factorization -- Directed trust -- Implicit feedback; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2021.107206
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: This paper focuses on improving the performance of recommender systems by the use of social trust information. Probabilistic matrix factorization is a classic algorithm for recommender systems. However, both the rating matrix and trust matrix become sparser, which makes the recommended results inaccurate. We propose a hybrid method based on probabilistic matrix factorization and directed trust. First, we apply the probabilistic matrix factorization approach to break down the trust matrix. Thus, the potential preferences of users, considering trusters and trustees, are obtained. This approach alleviates the problem of the sparsity of the trust matrix. Second, to capture the trust relations among users, we modify undirected trust to directed trust, since a user has different preferences when he is treated as a truster or trustee. Last, the two algorithms are combined to predict ratings. Experiments involving two datasets show that the proposed algorithm is superior to existing benchmark algorithms. Graphical abstract: Highlights: Differences between trusters and trustees are captured by directed trust in social influence strength. Trust matrix is decomposed to calculate the strength of the directed trust. Potential features of users' trust relationships are considered. Proposed algorithm outperformed benchmark algorithms and alleviated sparsity of the trust matrix.