Multi-view and multi-scale contrastive learning for graph collaborative filtering

doi:10.11772/j.issn.1001-9081.2024030393

Journal of Computer Applications ›› 2025, Vol. 45 ›› Issue (4): 1061-1068.DOI: 10.11772/j.issn.1001-9081.2024030393

• Artificial intelligence • Previous Articles Next Articles

Multi-view and multi-scale contrastive learning for graph collaborative filtering

Weichao DANG, Xinyu WEN(), Gaimei GAO, Chunxia LIU

College of Computer Science and Technology，Taiyuan University of Science and Technology，Taiyuan Shanxi 030024，China

Received:2024-04-08 Revised:2024-05-26 Accepted:2024-05-29 Online:2024-08-15 Published:2025-04-10
Contact: Xinyu WEN
About author:DANG Weichao， born in 1974， Ph. D.， associate professor. His research interests include intelligent computing， software reliability.
WEN Xinyu， born in 1998， M. S. candidate. His research interests include recommendation system.
GAO Gaimei， born in 1978， Ph. D.， associate professor. Her research interests include network security， cryptography.
LIU Chunxia， born in 1977， M. S.， associate professor. Her research interests include software engineering， database.
Supported by:
Shanxi Provincial Natural Science Foundation(202203021211194);Doctoral Research Start-up Fund of Taiyuan University of Science and Technology(20202063);Graduate Education Innovation Project of Taiyuan University of Science and Technology(SY2022063)

基于多视图多尺度对比学习的图协同过滤

党伟超, 温鑫瑜(), 高改梅, 刘春霞

太原科技大学计算机科学与技术学院，太原 030024

通讯作者: 温鑫瑜
作者简介:党伟超（1974—），男，山西运城人，副教授，博士，CCF会员，主要研究方向：智能计算、软件可靠性
温鑫瑜（1998—），男，山西晋中人，硕士研究生，主要研究方向：推荐系统
高改梅（1978—），女，山西吕梁人，副教授，博士，CCF会员，主要研究方向：网络安全、密码学
刘春霞（1977—），女，山西大同人，副教授，硕士，CCF会员，主要研究方向：软件工程、数据库。
基金资助:
山西省自然科学基金资助项目(202203021211194);太原科技大学博士科研启动基金资助项目(20202063);太原科技大学研究生教育创新项目(SY2022063)

Abstract

Abstract:

A Multi-View and Multi-Scale Contrastive Learning for graph collaborative filtering （MVMSCL） model was proposed to address the limitations of single view and the data sparsity in graph collaborative filtering recommendation methods. Firstly， an initial interaction diagram was constructed on the basis of user-item interactions， and multiple potential intentions in user-item interactions were considered to build multi-intention decomposition view. Secondly， the adjacency matrix was improved using high-order relationships to construct a collaborative neighbor view. Thirdly， the irrelevant noise interactions were removed to construct the adaptively enhanced initial interaction diagram and multi-intention decomposition view. Finally， contrastive learning paradigms with local， cross-layer， and global scales were introduced to generate self-supervised signals， thereby improving the recommendation performance. Experimental results on three public datasets， Gowalla， Amazon-book and Tmall， demonstrate that the recommendation performance of MVMSCL surpasses that of the comparison models. Compared with the optimal baseline model DCCF （Disentangled Contrastive Collaborative Filtering framework）， MVMSCL has the Recall@20 increased by 5.7%， 14.5% and 10.0%， respectively， and the Normalized Discounted Cumulative Gain NDCG@20 increased by 4.6%， 17.9% and 11.5%， respectively.

Key words: recommendation system, collaborative filtering, Graph Neural Network (GNN), multi-view, contrastive learning

摘要：

针对图协同过滤推荐方法存在的单一视图局限性和数据稀疏性问题，提出一种基于多视图多尺度对比学习的图协同过滤（MVMSCL）模型。首先，根据用户-项目交互构建初始交互图，并考虑用户-项目中存在的多种潜在意图，以构建多意图分解视图；其次，利用高阶关系改进邻接矩阵，以构建协同邻居视图；再次，去除不重要的噪声交互，以构建自适应增强的初始交互图和多意图分解视图；最后，引入局部、跨层和全局3种尺度的对比学习范式生成自监督信号，从而提高推荐性能。在Gowalla、Amazon-book和Tmall 3个公共数据集上的实验结果表明，MVMSCL的推荐性能均优于对比模型。与最优基线模型DCCF（Disentangled Contrastive Collaborative Filtering framework）相比，MVMSCL的召回率Recall@20分别提升了5.7%、14.5%和10.0%，归一化折损累计增益NDCG@20分别提升了4.6%、17.9%和11.5%。

关键词: 推荐系统, 协同过滤, 图神经网络, 多视图, 对比学习

CLC Number:

TP391.4

Weichao DANG, Xinyu WEN, Gaimei GAO, Chunxia LIU. Multi-view and multi-scale contrastive learning for graph collaborative filtering[J]. Journal of Computer Applications, 2025, 45(4): 1061-1068.

党伟超, 温鑫瑜, 高改梅, 刘春霞. 基于多视图多尺度对比学习的图协同过滤[J]. 《计算机应用》唯一官方网站, 2025, 45(4): 1061-1068.

Figures/Tables 9

Fig. 1 Construction of multiple views

Fig. 2 Pattern of contrastive learning

Tab. 1 Statistical information of experimental datasets

数据集	用户数	项目数	交互数	稀疏度/%
Gowalla	50 821	57 440	1 172 425	0.040
Amazon-book	78 578	77 801	2 240 156	0.037
Tmall	47 939	41 390	2 357 450	0.119

Tab. 2 Performance comparison of different models on public datasets

模型	Gowalla				Amazon-book				Tmall
模型	R@20	R@40	N@20	N@40	R@20	R@40	N@20	N@40	R@20	R@40	N@20	N@40
NCF	0.124 7	0.191 0	0.065 9	0.083 2	0.046 8	0.077 1	0.033 6	0.043 8	0.038 3	0.064 7	0.025 2	0.034 4
AutoR	0.140 9	0.214 2	0.071 6	0.090 5	0.054 6	0.091 4	0.035 4	0.048 2	0.033 6	0.061 1	0.020 3	0.029 5
NGCF	0.141 3	0.207 2	0.081 3	0.098 7	0.053 2	0.086 6	0.038 8	0.050 1	0.042 0	0.075 1	0.025 0	0.036 5
LightGCN	0.179 9	0.257 7	0.105 3	0.125 5	0.073 2	0.114 8	0.054 4	0.068 1	0.055 5	0.089 5	0.038 1	0.049 9
DGCF	0.178 4	0.251 5	0.106 9	0.125 9	0.068 8	0.107 3	0.051 3	0.064 0	0.054 4	0.086 7	0.037 2	0.048 4
DGCL	0.179 3	0.248 3	0.106 7	0.124 7	0.067 7	0.105 7	0.050 6	0.063 1	0.052 6	0.084 5	0.035 9	0.046 9
SGL-ED	0.180 9	0.255 9	0.106 7	0.126 2	0.077 4	0.120 4	0.057 8	0.071 9	0.057 4	0.091 9	0.039 3	0.051 3
SGL-ND	0.181 4	0.258 9	0.106 5	0.126 7	0.072 2	0.112 1	0.054 2	0.067 4	0.055 3	0.088 5	0.037 9	0.049 4
HCCF	0.181 8	0.260 1	0.106 1	0.126 5	0.082 4	0.128 2	0.062 5	0.077 6	0.062 3	0.098 6	0.042 5	0.055 2
LightGCL	0.182 5	0.260 1	0.107 7	0.128 0	0.083 6	0.128 0	0.064 3	0.079 0	0.063 2	0.097 1	0.044 4	0.056 2
DCCF	0.187 6	0.264 4	0.112 3	0.132 3	0.088 9	0.134 3	0.068 0	0.082 9	0.066 8	0.104 2	0.046 9	0.059 8
MVMSCL	0.198 3	0.274 2	0.117 5	0.137 4	0.101 8	0.148 9	0.080 2	0.099 5	0.073 5	0.113 2	0.052 3	0.066 0

Tab. 3 Comparison results of MVMSCL model’s ablation experiments

模型	R@20	R@40	N@20	N@40
MVMSCL- W	0.184 2	0.258 6	0.108 2	0.127 7
MVMSCL- $H α$	0.190 7	0.268 0	0.112 4	0.132 6
MVMSCL- $H β$	0.190 9	0.268 1	0.112 5	0.132 6
MVMSCL- T	0.187 7	0.264 4	0.112 1	0.132 9
MVMSCL-L	0.189 5	0.266 0	0.112 1	0.132 1
MVMSCL-G	0.195 8	0.272 7	0.116 1	0.136 3
MVMSCL-C	0.198 0	0.274 0	0.117 0	0.137 1
MVMSCL	0.198 3	0.274 2	0.117 5	0.137 4

Tab. 3 Comparison results of MVMSCL model’s ablation experiments

模型	R@20	R@40	N@20	N@40
MVMSCL- W	0.184 2	0.258 6	0.108 2	0.127 7
MVMSCL- $H α$	0.190 7	0.268 0	0.112 4	0.132 6
MVMSCL- $H β$	0.190 9	0.268 1	0.112 5	0.132 6
MVMSCL- T	0.187 7	0.264 4	0.112 1	0.132 9
MVMSCL-L	0.189 5	0.266 0	0.112 1	0.132 1
MVMSCL-G	0.195 8	0.272 7	0.116 1	0.136 3
MVMSCL-C	0.198 0	0.274 0	0.117 0	0.137 1
MVMSCL	0.198 3	0.274 2	0.117 5	0.137 4

Fig. 3 Influence of parameter λ1 on recommendation performance of model

Fig. 4 Influence of layers of GNN on performance of model

Tab. 4 Experimental results of multi-view fusion strategies

模型	R@20	R@40	N@20	N@40
$Z$	0.136 6	0.198 7	0.081 4	0.097 6
$Z + W + H α$	0.172 3	0.245 5	0.101 7	0.121 1
$Z + W + H β$	0.171 9	0.245 1	0.101 4	0.120 5
$Z + W + T$	0.166 3	0.241 1	0.097 2	0.116 6
$Z + H α + T$	0.184 2	0.258 6	0.108 2	0.127 2

Tab. 4 Experimental results of multi-view fusion strategies

模型	R@20	R@40	N@20	N@40
$Z$	0.136 6	0.198 7	0.081 4	0.097 6
$Z + W + H α$	0.172 3	0.245 5	0.101 7	0.121 1
$Z + W + H β$	0.171 9	0.245 1	0.101 4	0.120 5
$Z + W + T$	0.166 3	0.241 1	0.097 2	0.116 6
$Z + H α + T$	0.184 2	0.258 6	0.108 2	0.127 2

Tab. 5 Multi-scale comparative strategies

模型	R@20	R@40	N@20	N@40
AllLocal+Layer	0.180 9	0.106 8	0.256 2	0.126 5
AllGlobal+Layer	0.192 1	0.114 6	0.269 5	0.134 8
Local+Global+Layer	0.198 3	0.117 5	0.274 2	0.137 4

References 21

1	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）： No.3.
2	PARK S， YOON M， LEE J W， et al. Toward a better understanding of loss functions for collaborative filtering ［C］// Proceedings of the 32nd ACM International Conference on Information and Knowledge Management. New York： ACM， 2023： 2034-2043.
3	WANG X， HE X， WANG M， et al. Neural graph collaborative filtering ［C］// Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2019： 165-174.
4	HE X， DENG K， WANG X， et al. LightGCN： simplifying and powering graph convolution network for recommendation ［C］// Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2020： 639-648.
5	MAO K， ZHU J， XIAO X， et al. UltraGCN： ultra simplification of graph convolutional networks for recommendation ［C］// Proceedings of the 30th ACM International Conference on Information and Knowledge Management. New York： ACM， 2021： 1253-1262.
6	LINSKER R. Self-organization in a perceptual network ［J］. Computer， 1988， 21（3）：105-117.
7	杨兴耀，陈羽，于炯，等. 结合自我特征和对比学习的推荐模型［J］. 计算机应用， 2024， 44（9）：2704-2710.
	YANG X Y， CHEN Y， YU J， et al. Recommendation model combining self-features and contrastive learning ［J］. Journal of Computer Applications， 2024， 44（9）：2704-2710.
8	WU J， WANG X， FENG F， et al. Self-supervised graph learning for recommendation ［C］// Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2021： 726-735.
9	LIN Z， TIAN C， HOU Y， et al. Improving graph collaborative filtering with neighborhood-enriched contrastive learning ［C］// Proceedings of the ACM Web Conference 2022. New York： ACM， 2022： 2320-2329.
10	XIA L， HUANG C， XU Y， et al. Hypergraph contrastive collaborative filtering ［C］// Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2022： 70-79.
11	WANG W， FENG F， HE X， et al. Deconfounded recommendation for alleviating bias amplification ［C］// Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York： ACM， 2021： 1717-1725.
12	ABRAMOVICH S， PERSSON L E. Some new estimates of the ‘Jensen gap’ ［J］. Journal of Inequalities and Applications， 2016， 2016： No.39.
13	GAO X， SITHARAM M， ROITBERG A E. Bounds on the Jensen gap， and implications for mean-concentrated distributions ［J］. Information Processing and Management， 2019， 16（2）： No.14.
14	HAMERS L， HEMERYCK Y， HERWEYERS G， et al. Similarity measures in scientometric research： the Jaccard index versus Salton’s cosine formula ［J］. Information Processing and Management， 1989， 25（3）： 315-318.
15	TIAN C， XIE Y， LI Y， et al. Learning to denoise unreliable interactions for graph collaborative filtering ［C］// Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2022： 122-132.
16	REN X， XIA L， ZHAO J， et al. Disentangled contrastive collaborative filtering ［C］// Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2023：1137-1146.
17	CAI X， HUANG C， XIA L， et al. LightGCL： simple yet effective graph contrastive learning for recommendation ［EB/OL］. ［2024-05-23］. .
18	HE X， LIAO L， ZHANG H， et al. Neural collaborative filtering［C］// Proceedings of the 26th International Conference on World Wide Web. Republic and Canton of Geneva： International World Wide Web Conferences Steering Committee， 2017： 173-182.
19	SEDHAIN S， MENON A K， SANNER S， et al. AutoRec： autoencoders meet collaborative filtering ［C］// Proceedings of the 24th International Conference on World Wide Web. New York： ACM， 2015： 111-112.
20	WANG X， JIN H， ZHANG A， et al. Disentangled graph collaborative filtering ［C］// Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. New York： ACM， 2020： 1001-1010.
21	LI H， WANG X， ZHANG Z， et al. Disentangled contrastive learning on graphs ［C］// Proceedings of the 35th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2021： 21872-21884.

[1]	Lan YOU, Yuang ZHANG, Yuan LIU, Zhijun CHEN, Wei WANG, Xing ZENG, Zhangwei HE. Developer recommendation for open-source projects based on collaborative contribution network [J]. Journal of Computer Applications, 2025, 45(4): 1213-1222.
[2]	Cong WANG, Yancui SHI. Group recommendation model by graph neural network based on multi-perspective learning [J]. Journal of Computer Applications, 2025, 45(4): 1205-1212.
[3]	Guangju YANG, Tianjian LUO, Kaijun WANG, Siqi YANG. Multi-branch multi-view based contextual contrastive representation learning method for time series [J]. Journal of Computer Applications, 2025, 45(4): 1042-1052.
[4]	Renjie TIAN, Mingli JING, Long JIAO, Fei WANG. Recommendation algorithm of graph contrastive learning based on hybrid negative sampling [J]. Journal of Computer Applications, 2025, 45(4): 1053-1060.
[5]	Yuanlong WANG, Tinghua LIU, Hu ZHANG. Commonsense question answering model based on cross-modal contrastive learning [J]. Journal of Computer Applications, 2025, 45(3): 732-738.
[6]	Wei CHEN, Changyong SHI, Chuanxiang MA. Crop disease recognition method based on multi-modal data fusion [J]. Journal of Computer Applications, 2025, 45(3): 840-848.
[7]	Sheng YANG, Yan LI. Contrastive knowledge distillation method for object detection [J]. Journal of Computer Applications, 2025, 45(2): 354-361.
[8]	Handa MA, Yadong WU. Multi-domain spatiotemporal hierarchical graph neural network for air quality prediction [J]. Journal of Computer Applications, 2025, 45(2): 444-452.
[9]	Qijian CAI, Wei TAN. Semantic graph enhanced multi-modal recommendation algorithm [J]. Journal of Computer Applications, 2025, 45(2): 421-427.
[10]	Xiaosheng YU, Zhixin WANG. Sequential recommendation model based on multi-level graph contrastive learning [J]. Journal of Computer Applications, 2025, 45(1): 106-114.
[11]	Zidong CHENG, Peng LI, Feng ZHU. Potential relation mining in internet of things threat intelligence knowledge graph [J]. Journal of Computer Applications, 2025, 45(1): 24-31.
[12]	Wenbo ZHAO, Zitong MA, Zhe YANG. Link prediction model based on directed hypergraph adaptive convolution [J]. Journal of Computer Applications, 2025, 45(1): 15-23.
[13]	Zhuoyue OU, Xiuqin DENG, Lei CHEN. Self-adaptive multi-view clustering algorithm with complementarity based on weighted anchors [J]. Journal of Computer Applications, 2025, 45(1): 115-126.
[14]	Tingjie TANG, Jiajin HUANG, Jin QIN. Session-based recommendation with graph auxiliary learning [J]. Journal of Computer Applications, 2024, 44(9): 2711-2718.
[15]	Xingyao YANG, Yu CHEN, Jiong YU, Zulian ZHANG, Jiaying CHEN, Dongxiao WANG. Recommendation model combining self-features and contrastive learning [J]. Journal of Computer Applications, 2024, 44(9): 2704-2710.

Multi-view and multi-scale contrastive learning for graph collaborative filtering

基于多视图多尺度对比学习的图协同过滤

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 9

References 21

Related Articles 15

Recommended Articles

Metrics