Adaptive social recommendation combined with the multi-domain influence. Issue 113 (January 2023)

• Record Type: Journal Article
• Authors: Qian, Fulan; Qin, Kaili; Chen, Hai; Chen, Jie; Zhao, Shu; Zhang, Yanping
• Is Part Of: Information systems. Issue 113(2023)
• Publication Date: 2023-01
• Subjects: Recommender system -- Social influence -- Attention mechanism; Database management -- Periodicals; Electronic data processing -- Periodicals; Bases de données -- Gestion -- Périodiques; Informatique -- Périodiques; Database management; Electronic data processing; Periodicals; 005.7
• DOI: 10.1016/j.is.2022.102145
• Languages: English
• ISSNs: 0306-4379

Abstract

Abstract: Social relationships help to model user's potential preferences and improve recommendation accuracy. In social recommendation, user decision-making will be affected by his own historical interaction items and social friends. Most social recommendations consider these two aspects separately, but users are also driven by both when making decisions. To address the above issue, this paper proposes an adaptive social recommendation method based on attention mechanism, Adaptive Social Recommendation combine with Multi-domain influence (ASRM). Specifically, we defined three domains, item domain, social domain, and common domain to measure the possible impact of social relations and item interactions on users. We propose an adaptive attention module to simulate users being influenced by social relations and historical interaction items in these three domains, and obtain user's potential preferences. In addition, we jointly optimize by calculating the loss of each domain, resulting in more accurate recommendations. A large number of experiments on three real datasets have demonstrated the effectiveness of our method. The code is available at https://github.com/qinkaili/ASRM . Highlights: A new multi-domain adaptive social recommendation method. Considers the combined impact of historical items and social information. Item, social, and common domain are used to simulate item and social impact on users. Using attention mechanism to distinguish the importance of domain information.