Global-mirror graph network for session-based recommendation

Session-based recommendation(SBR) is a hot research direction, and how to solve the problem of sparse data in sessions is a challenging task. Although the existing model mirror graph enhanced neural model for session-based recommendation(MGS) solves the problem of data sparsity by introducing additional information from items, this model is still unable to model the general interest of the session well due to the limitations of data attribute information and the interference of noise, resulting in a certain amount of room for improvement in recommendation accuracy. To address this issue, this paper proposes a Global-Mirror Graph Model for Session-based Recommendation (GMGS), based on the research work related to MGS models, combining global graph modeling and reverse position awareness. Specifically, the GMGS model first builds a global graph based on all item transformations, and uses this to capture contextual information between sessions, thereby completing the modeling of general session interests. Then, a gating mechanism is used to combine reverse position information and the last click to obtain a session representation for final prediction. Finally, we conducted a large number of comparative experiments on three real datasets, and the GMGS model proposed in this paper achieved better results than the current representative recommendation models. Compared with the MGS model, the MRR@20 indicator has increased by 2.6%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document}, 0.6%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document}, and 6.5%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document} on the three datasets, respectively. In addition, we also conducted ablation experiments on three datasets to verify the effectiveness of each component improvement.