基于图注意力网络的群推荐方法

doi:10.12305/j.issn.1001-506X.2026.02.16

系统工程与电子技术 ›› 2026, Vol. 48 ›› Issue (2): 545-555.doi: 10.12305/j.issn.1001-506X.2026.02.16

• 系统工程 • 上一篇

基于图注意力网络的群推荐方法

王亚楠¹(), 梁如霞¹^,², 王晓康¹^,³^,*, 柳叶¹, 王坚强¹()

1. 中南大学商学院，湖南长沙 410083
2. 华中师范大学计算机学院，湖北武汉 430079
3. 深圳大学管理学院，广东深圳 518060

收稿日期:2024-10-18 修回日期:2025-05-28 出版日期:2025-05-20 发布日期:2025-05-20
通讯作者: 王晓康 E-mail:201611122@csu.edu.cn;jqwang@csu.edu.cn
作者简介:王亚楠（1998—），女，博士研究生，主要研究方向为大数据分析与人工智能、辅助决策
梁如霞（1992—），女，助理研究员，博士，主要研究方向为智能教育、跨域推荐、群推荐、迁移学习
柳　叶（1996—），女，博士研究生，主要研究方向为数据挖掘、深度学习
王坚强（1963—），男，教授，博士，主要研究方向为决策理论与方法、大数据分析与机器学习
基金资助:
湖南省社会科学评审委员会重点项目（XSP18ZDI021）；湖南省研究生科研创新项目(CX20230132)资助课题

Group recommendation method based on graph attention network

Yanan WANG¹(), Ruxia LIANG¹^,², Xiaokang WANG¹^,³^,*, Ye LIU¹, Jianqiang WANG¹()

1. School of Business，Central South University，Changsha 410073，China
2. School of Computer Science，Central China Normal University，Wuhan 430079，China
3. College of Management，Shenzhen University，Shenzhen 518060，China

Received:2024-10-18 Revised:2025-05-28 Online:2025-05-20 Published:2025-05-20
Contact: Xiaokang WANG E-mail:201611122@csu.edu.cn;jqwang@csu.edu.cn

摘要/Abstract

摘要：

针对群推荐系统的数据稀疏性挑战，以及现有群推荐方法忽略群、群成员以及不同备选物品之间的复杂关联关系问题，提出一种基于图注意力网络的群推荐方法（group recommendation method based on graph attention network，GAT-GRM）。首先，将群推荐系统中群、群成员以及不同物品间的复杂关联刻画为层次关系图数据，包括用户?物品交互关系图、群组?用户包含关系图、群组?物品交互关系图等。其次，利用图注意力网络内在地聚合各类交互关系图，从历史交互数据中动态学习群偏好、用户偏好和物品特征。最后，基于群偏好、用户偏好和物品特征进行物品评分预测。实验结果表明，在CAMRa2011数据集上，GAT-GRM的性能显著优于各类基准算法。在稀疏度98.89%的群推荐任务下，GAT-GRM平均绝对偏差和均方根误差相较最优基准算法分别降低9.3%和9.6%。

关键词: 推荐系统, 群推荐系统, 聚合策略, 图神经网络

Abstract:

Existing group recommendation methods neglect the complex associations among groups, group members, and diverse candidate items, and face challenges related to data sparsity. To address the above issues, this paper proposes a group recommendation method based on graph attention network (GAT-GRM). Firstly, the complex associations among groups, group members, and different items in the group recommendation system are characterized as hierarchical graph data, including user-item interaction graphs, group-user inclusion graphs, and group-item interaction graphs. Secondly, graph attention networks are employed to aggregate various types of interaction graphs to dynamically learn group preferences, user preferences and item characteristics from historical interaction data. Finally, item score prediction is performed based on group preferences, user preferences, and item features. Experiments conducted on the CAMRa2011 dataset demonstrate that the performance of GAT-GRM is significantly superior to that of all baseline algorithms. For the group recommendation task with 98.89% sparsity, the mean absolute error and root mean square error of GAT-GRM are reduced by 9.3% and 9.6% respectively compared to the optimal baseline algorithm.

Key words: recommender system, group recommender system, aggregation strategy, graph neural networks

中图分类号:

TP 18

王亚楠, 梁如霞, 王晓康, 柳叶, 王坚强. 基于图注意力网络的群推荐方法[J]. 系统工程与电子技术, 2026, 48(2): 545-555.

Yanan WANG, Ruxia LIANG, Xiaokang WANG, Ye LIU, Jianqiang WANG. Group recommendation method based on graph attention network[J]. Systems Engineering and Electronics, 2026, 48(2): 545-555.

图/表 10

表1

图1

图2

表2

表3

表4

表5

表6

表7

图3

参考文献 38

1	吴国栋, 查志康, 涂立静, 等. 图神经网络推荐研究进展[J]. 智能系统学报, 2020, 15 (1): 14- 24.
	WU G D, ZHA Z K, TU L J, et al. Research advances in graph neural network recommendation[J]. CAAI Transactions on Intelligent Systems, 2020, 15 (1): 14- 24.
2	GAO C, ZHENG Y, LI N, et al. A survey of graph neural networks for recommender systems: challenges, methods, and directions[J]. ACM Transactions on Recommender Systems, 2023, 1 (1): 237940542.
3	VELIČKOVIĆ P. Everything is connected: graph neural networks[J]. Current Opinion in Structural Biology, 2023, 79, 102538.
4	ZHANG H, WU B, YUAN X L, et al. Trustworthy graph neural networks: aspects, methods, and trends[J]. Proceedings of the IEEE, 2024, 112 (2): 97- 139.
5	SHARMA K, LEE Y C, NAMBI S, et al. A survey of graph neural networks for social recommender systems[J]. ACM Computing Surveys, 2024, 56 (10): 1- 34.
6	MA W Z, WANG Y X, ZHU Y M, et al. MADM: a model-agnostic denoising module for graph-based social recommendation[C]// Proc. of the 17th ACM International Conference on Web Search and Data Mining, 2024: 501−509.
7	FAN W Q, MA Y, LI Q, et al. Graph neural networks for social recommendation[C]// Proc. of the World Wide Web Conference: Association for Computing Machinery, 2019: 417−426.
8	FAN W Q, MA Y, LI Q, et al. A graph neural network framework for social recommendations[J]. IEEE Trans. on Knowledge and Data Engineering, 2020, 34 (5): 2033- 2047.
9	HUANG C, XU H, XU Y, et al. Knowledge-aware coupled graph neural network for social recommendation[C]// Proc. of the AAAI Conference on Artificial Intelligence, 2021, 35(5): 4115−4122.
10	CHEN H X, YIN H Z, CHEN T, et al. Social boosted recommendation with folded bipartite network embedding[J]. IEEE Trans. on Knowledge and Data Engineering, 2022, 34 (2): 914- 926. doi: 10.1109/TKDE.2020.2982878
11	CHEN H, BEI Y C, HUANG W B, et al. Graph cross-correlated network for recommendation[J]. IEEE Trans. on Knowledge and Data Engineering, 2025, 37 (2): 710- 723. doi: 10.1109/TKDE.2024.3491778
12	SHLOMO B, JILL F. Group-based recipe recommendations: analysis of data aggregation strategies[C]// Proc. of the 4th ACM Conference on Recommender Systems: Association for Computing Machinery, 2010: 111−118.
13	SIHEM A Y, ROY S B, CHAWLAT A, et al. Group recommendation: semantics and efficiency[J]. Proceedings of the VLDB Endowment, 2009, 2 (1): 754- 765.
14	ELISA Q, EMANUELE R, LETIZIA T. Recommending new items to ephemeral groups using contextual user influence[C]// Proc. of the 10th ACM Conference on Recommender Systems, 2016: 285−292.
15	HUANG T C K. Recommendations of closed consensus temporal patterns by group decision making[J]. Knowledge-Based Systems, 2013, 54, 318- 328. doi: 10.1016/j.knosys.2013.10.003
16	WANG W, ZHANG G Q, LU J. Member contribution-based group recommender system[J]. Decision Support Systems, 2016, 87, 80- 93. doi: 10.1016/j.dss.2016.05.002
17	SEKO S, YAGI T, MOTEGI M, et al. Group recommendation using feature space representing behavioral tendency and power balance among members[C]// Proc. of the 5th ACM Conference on Recommender Systems, 2011: 101−108.
18	QIN D, ZHOU X M, CHEN L, et al. Dynamic connection-based social group recommendation[J]. IEEE Trans. on Knowledge and Data Engineering, 2020, 32 (3): 453- 467. doi: 10.1109/TKDE.2018.2879658
19	DU Y L, MENG X W, ZHANG Y J. CVTM: a content-venue-aware topic model for group event recommendation[J]. IEEE Trans. on Knowledge and Data Engineering, 2019, 32 (7): 1290- 1303. doi: 10.1109/tkde.2019.2904066
20	WANG X M, LIU Y, LU J, et al. TruGRC: trust-aware group recommendation with virtual coordinators[J]. Future Generation Computer Systems, 2019, 94, 224- 236. doi: 10.1016/j.future.2018.11.030
21	魏玲, 权晨雪. 融合多维特征与兴趣漂移的虚拟学术社区群推荐模型[J]. 现代情报, 2023, 43 (7): 48- 63. doi: 10.3969/j.issn.1008-0821.2023.07.006
	WEI L, QUAN C X. A group recommendation model for virtual academic community group recommendation model integrating multidimensional features and interest drift[J]. Journal of Modern Information, 2023, 43 (7): 48- 63. doi: 10.3969/j.issn.1008-0821.2023.07.006
22	ALATIYYAH M. A novel group tour trip recommender model for personalized travel systems[J]. PeerJ Computer Science, 2025, 11, e2589. doi: 10.7717/peerj-cs.2589
23	CAO D, HE X N, MIAO L H, et al. Attentive group recommendation[C]// Proc. of the 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, 2018: 645−654.
24	JIANG W B, SUN Y R. Social-RippleNet: jointly modeling of ripple net and social information for recommendation[J]. Applied Intelligence, 2023, 53 (3): 3472- 3487. doi: 10.1007/s10489-022-03620-2
25	QIAO P P, ZHANG Z W, LI Z T, et al. TAG: joint triple-hierarchical attention and GCN for review-based social recommender system[J]. IEEE Trans. on Knowledge and Data Engineering, 2023, 35 (10): 9904- 9919. doi: 10.1109/TKDE.2022.3194952
26	YIN H Z, WANG Q Y, ZHENG K, et al. Social influence-based group representation learning for group recommendation[C]// Proc. of the IEEE 35th International Conference on Data Engineering, 2019: 566−577.
27	YIN H Z, WANG Q Y, ZHENG K, et al. Overcoming data sparsity in group recommendation[J]. IEEE Trans. on Knowledge and Data Engineering, 2020, 34 (5): 3447- 3460. doi: 10.1109/tkde.2020.3023787
28	LI G H, CHEN Q, ZHENG B, et al. Group-based recurrent neural networks for POI recommendation[J]. ACM/IMS Transactions on Data Science, 2020, 1 (1): 1- 18.
29	GUO L, YIN H Z, CHEN T, et al. Hierarchical hyperedge embedding-based representation learning for group recommendation[J]. ACM Transactions on Information Systems, 2020, 40 (1): 1- 27.
30	YU L, LENG Y F, ZHANG D S, et al. Collaborative group embedding and decision aggregation based on attentive influence of individual members: a group recommendation perspective[J]. Decision Support Systems, 2023, 165, 113894. doi: 10.1016/j.dss.2022.113894
31	LENG Y F, YU L. Incorporating global and local social networks for group recommendations[J]. Pattern Recognition, 2022, 127, 108601. doi: 10.1016/j.patcog.2022.108601
32	叶佳鑫, 熊回香, 易明, 等. 融合影响力传播的社交网络群推荐方法[J]. 情报学报, 2022, 41 (4): 364- 374. doi: 10.3772/j.issn.1000-0135.2022.04.004
	YE J X, XIONG H X, YI M, et al. Group recommendation in social networks based on influence spread[J]. Journal of the China Society for Scientific and Technical Information, 2022, 41 (4): 364- 374. doi: 10.3772/j.issn.1000-0135.2022.04.004
33	HONG T, IBRAHIM N F. SoLGR: social enhancement group recommendation via light graph convolution networks[J]. IEEE Access, 2023, 11, 74828- 74838. doi: 10.1109/ACCESS.2023.3280629
34	ZHANG Q, LIAO W H, ZHANG G Q, et al. A deep dual adversarial network for cross-domain recommendation[J]. IEEE Trans. on Knowledge and Data Engineering, 2021, 35 (4): 3266- 3278. doi: 10.1109/tkde.2021.3132953
35	WANG P P, LI L, XIE Q, et al. Social dual-effect driven group modeling for neural group recommendation[J]. Neurocomputing, 2022, 481, 258- 269. doi: 10.1016/j.neucom.2022.01.093
36	KOREN Y. Factorization meets the neighborhood: a multifaceted collaborative filtering model[C]// Proc. of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2008: 426−434.
37	ORTEGA F, HERNANDO A, BOBADILLA J, et al. Recommending items to group of users using matrix factorization based collaborative filtering[J]. Information Sciences, 2016, 345, 313- 324. doi: 10.1016/j.ins.2016.01.083
38	LIANG R X, ZHANG Q, LU J, et al. A cross-domain group recommender system with a generalized aggregation strategy[C]// Proc. of the Developments of Artificial Intelligence Technologies in Computation and Robotics, 2020: 455−462.

数据类型	文献方法	群组类型	用户偏好	物品特征	聚合策略	推荐方法	应用场景
显式评分	[17]	熟人	固定	固定	均值策略	邻域推荐	影视推荐
	[16]	随机	固定	固定	加权策略	邻域推荐	影视推荐
	[20]	熟人	固定	固定	基于信任聚合	矩阵分解	社会化推荐
	[18]	随机	固定	固定	考虑贡献和积极性的加权聚合	矩阵分解	在线社群推荐
隐式交互	[23]	熟人	固定	固定	动态加权	神经协同过滤框架	旅游推荐
	[19]	随机	固定	固定	加权聚合	概率主题模型	活动推荐
	[26−28]	随机	固定	固定	自注意力机制	图卷积网络	社会化推荐
	[35]	熟人/ 随机	固定	固定	注意力机制	神经群推荐框架	社会化推荐
	[31]	熟人	不固定	固定	注意力机制	神经协同过滤框架	社会化推荐
	[32]	熟人	不固定	固定	基于影响力的加权聚合	社交网络分析	影视和旅游推荐
	[21]	熟人	不固定	不固定	动态加权	神经协同过滤框架	学术社区推荐
	[30]	—	固定	固定	注意力机制	GNN	社会化推荐
	[33]	熟人	不固定	固定	基于图结构的聚合	图卷积网络	社会化推荐
	[22]	—	固定	固定	动态聚合	蚁群优化算法	旅游推荐

统计项	原始数据集	数据集A	数据集B	数据集C
用户/群组数量	1644	289	289	289
物品数量	7792	1701	1701	1701
评分数量	146408	5445	3925	1965
稀疏度/%	98.86	98.89	99.20	99.60
评分均值	3.6838	2.1306	2.1803	3.8124

模型	数据集A	数据集B	数据集C
Bias-SVD	0.4956±0.0127	0.4705±0.0112	0.6236±0.0205
MFCFGR	0.4855±0.0101	0.5456±0.0131	0.6260±0.0289
CDGRS-A	0.4348±0.0155	0.4618±0.0175	0.5475±0.0368
CDGRS-MS	0.4346±0.0165	0.4611±0.0176	0.5487±0.0366
CDGRS-LM	0.4344±0.0149	0.4619±0.0165	0.5455±0.0361
GAT-GRM	0.3941±0.0135	0.4081±0.0153	0.4844±0.0340

模型	数据集A	数据集B	数据集C
Bias-SVD	0.6136±0.0115	0.5860±0.0099	0.7977±0.0221
MFCFGR	0.6315±0.0122	0.6992±0.0138	0.8341±0.0354
CDGRS-A	0.5409±0.0130	0.5591±0.0180	0.6988±0.0336
CDGRS-MS	0.5395±0.0133	0.5579$ \pm $0.0176	0.6969±0.0330
CDGRS-LM	0.5400±0.0125	0.5591$ \pm $0.0172	0.6985±0.0347
GAT-GRM	0.4880±0.0140	0.5274$ \pm $0.0181	0.6472±0.0349

模型	数据集A	数据集B	数据集C
Bias-SVD	0.6549±0.0080	0.6549±0.0080	0.6549±0.0080
CDGRS-A	1.0145±0.0117	1.0145±0.0117	1.0145±0.0117
CDGRS-MS	1.0145±0.0117	1.0145±0.0117	1.0145±0.0117
CDGRS-LM	1.0145±0.0117	1.0145±0.0117	1.0145±0.0117
GAT-GRM	0.5966±0.0054	0.6091±0.0118	0.6064±0.0026

基于图注意力网络的群推荐方法

Group recommendation method based on graph attention network

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 38

相关文章 2

编辑推荐

Metrics

本文评价

指标	方法	数据集A	数据集B	数据集C
MAE	GAT-GRM-w/o M	0.4121±0.0121	0.4167±0.0087	0.4942±0.0094
	GAT-GRM-w/o I	0.4090±0.0122	0.4162±.0155	0.4910±0.0131
	GAT-GRM-w/o O	0.4204±0.0158	0.4183±0.0172	0.4967±0.0126
	GAT-GRM	0.3941±0.0135	0.4081±0.0153	0.4844±0.0340
RMSE	GAT-GRM-w/o M	0.4988±0.0088	0.5361±0.0098	0.6489±0.0079
	GAT-GRM-w/o I	0.4990±0.0048	0.5376±0.0148	0.6521±0.0171
	GAT-GRM-w/o O	0.5018±0.0140	0.5384±0.0128	0.6591±0.0013
	GAT-GRM	0.4880±0.0140	0.5274±0.0181	0.6472±0.0349

[1]	陈凯, 张德平. 基于图神经网络的导弹时序规划方法[J]. 系统工程与电子技术, 2025, 47(3): 862-870.
[2]	樊成, 王布宏, 田继伟. 基于多任务学习图卷积模型的航空网络节点分类[J]. 系统工程与电子技术, 2022, 44(7): 2341-2349.

模型	数据集A	数据集B	数据集C
Bias-SVD	0.9030±0.0092	0.9030±0.0092	0.9030±0.0092
CDGRS-A	1.3428±0.0113	1.3428±0.0113	1.3428±0.0113
CDGRS-MS	1.3428±0.0113	1.3428±0.0113	1.3428±0.0113
CDGRS-LM	1.3428±0.0113	1.3428±0.0113	1.3428±0.0113
GAT-GRM	0.8510±0.0089	0.8508±0.0063	0.8505±0.0053