基于时间和关系感知的图协同过滤跨域序列推荐

任豪; 刘柏嵩; 孙金杨; 董倩; 钱江波

doi:10.7544/issn1000-1239.202110545

基于时间和关系感知的图协同过滤跨域序列推荐

宁波大学信息科学与工程学院　浙江宁波　315211

基金项目: 国家自然科学基金项目（62271274）；浙江省自然科学基金项目（LZ20F020001）；宁波市2025重大专项科研项目（20211ZDYF020036）；宁波市自然科学基金项目（2021J091）

详细信息

作者简介:
任豪: 1994年生.硕士.CCF学生会员.主要研究方向为跨域推荐、序列推荐、图神经网络

刘柏嵩: 1971年生.博士. CCF专业会员.主要研究方向为信息检索与推荐系统、多方可信计算与隐私保护、大数据分析与知识组织

孙金杨: 1995年生.硕士.CCF学生会员.主要研究方向为推荐系统、数据挖掘和深度学习

董倩: 1997年生.硕士研究生.CCF学生会员.主要研究方向为推荐系统、联邦学习和机器学习

钱江波: 1974年生.博士.CCF高级会员.主要研究方向为数据库管理、流数据处理、多维索引和布鲁姆过滤器

通讯作者:
刘柏嵩（lbs@nbu.edu.cn）

中图分类号: TP391
计量
- 文章访问数: 307
- HTML全文浏览量: 44
- PDF下载量: 176
出版历程
- 收稿日期: 2021-06-02
- 修回日期: 2022-06-06
- 网络出版日期: 2023-02-10
- 刊出日期: 2022-12-31

A Time and Relation-Aware Graph Collaborative Filtering for Cross-Domain Sequential Recommendation

Faculty of Electrical Engineering and Computer Science, Ningbo University, Ningbo, Zhejiang 315211

Funds: This work was supported by the National Natural Science Foundation of China (62271274), the Natural Science Foundation of Zhejiang Province (LZ20F020001), the 2025 Major Special Scientific Research Project of Ningbo City (20211ZDYF020036), and the Natural Science Foundation of Ningbo City (2021J091).

摘要

摘要:
跨域序列推荐旨在从给定的某用户在不同领域中的历史交互序列中挖掘其偏好，预测其在多个领域中最可能与之交互的下一个项目，以缓解数据稀疏对用户意图捕捉和预测的影响. 受协同过滤思想启发，提出一种基于时间和关系感知的图协同过滤跨域序列推荐（time and relation-aware graph collaborative filtering for cross-domain sequential recommendation, TRaGCF）算法，充分挖掘用户高阶行为模式同时利用跨域用户行为模式双向迁移，解决序列推荐中的数据稀疏问题. 首先，为获得用户行为序列中项目间复杂的时序依赖关系，提出时间感知图注意力（time-aware graph attention, Ta-GAT）学习项目的域间序列级表示；其次，通过域内用户-项目交互二部图挖掘用户的行为偏好，提出关系感知图注意力（relation-aware graph attention, Ra-GAT）学习项目协同表示和用户协同偏好表示，为用户偏好特征的跨域迁移提供基础；最后为同步提高2个领域中的推荐效果，提出用户偏好特征双向迁移模块（user preference feature bi-directional transfer module, PBT），实现迁移用户域间共有偏好，保留用户域内特有偏好. 在Amazon Movie-Book和Food-Kitchen数据集上验证了算法的正确性和有效性. 实验结果表明，在跨域序列推荐场景下考虑项目间深层复杂的关联关系对挖掘用户意图十分必要；实验还验证了在跨域迁移用户偏好过程中保留域内用户特有偏好对全面用户画像的重要性.
- 跨域序列推荐 /
- 图协同过滤 /
- 时间感知图注意力机制 /
- 关系感知注意力机制 /
- 数据稀疏
Abstract:
Cross-domain sequential recommendation aims to mine a given user’s preferences from the historical interaction sequences in different domains and to predict the next item that the user is most likely to interact with among multiple domains, further to mitigate the impact of data sparsity on the capture and prediction for users’ intents. Inspired by the idea of collaborative filtering, a time and relation-aware graph collaborative filtering for cross-domain sequential recommendation (TRaGCF) algorithm is proposed to solve the problem of data sparsity by uncovering users’ high-order behavior patterns as well as utilizing the characteristics of bi-directional migration of user behavior patterns across domains. Firstly, we propose a time-aware graph attention (Ta-GAT) mechanism to obtain the cross-domain sequence-level item representation. Then, a user-item interaction bipartite graph in the domain is used to mine users’ preferences, and a relation-aware graph attention (Ta-GAT) mechanism is proposed to learn item collaborative representation and user collaborative representation, which creates the foundation for cross-domain transfer of user preferences. Finally, to simultaneously improve the recommendation results in both domains, a user preference feature bi-directional transfer module (PBT) is proposed, transferring shared user preferences across domains and retaining specific preferences within one domain. The accuracy and effectiveness of our model are validated by two experimental datasets, Amazon Movie-Book and Food-Kitchen. The experimental results have demonstrated the necessity of considering intricate correlations between items in a cross-domain sequential recommendation scenario for mining users’ intents, and the results also prove the importance of preserving users’ specific preferences in creating a comprehensive user portrait when transferring users’ preferences across domains.
- cross-domain sequential recommendation /
- graph collaborative filter /
- time-aware graph attention mechanism /
- relation-aware graph attention mechanism /
- data sparsity

HTML全文

图 1 TRaGCF模型框架

Figure 1. The frame of TRaGCF

下载: 全尺寸图片幻灯片

图 2 时间感知序列图

Figure 2. Time-aware sequential graph

下载: 全尺寸图片幻灯片

图 3 用户−项目交互二部图

Figure 3. A user-item interaction bipartite graph

下载: 全尺寸图片幻灯片

图 4 跨域交互图

Figure 4. Cross-domain interaction graph

下载: 全尺寸图片幻灯片

图 5 $\lambda$ 对模型性能的影响

Figure 5. The impact of $\lambda$ on the performance of the model

下载: 全尺寸图片幻灯片

表 1 预处理之后的数据集统计分析

Table 1 The Statistics of the Datasets After Preprocessing

数据集	项目总数	测试集项目数	训练集项目数	验证集项目数	平均序列长度
Movie-Book	36845/63937	34732	19861	9274	11.98
Food-Kitchen	29207/34886	25766	17280	7650	9.91
注：Movie-Book的项目总数36845/63937表示在Movie域中项目总数为36845，Book域中的项目总数为63937；Food-Kitchen同理.

下载: 导出CSV

表 2 在Amazon Movie-Book上的实验结果

Table 2 Experimental Results on Amazon Movie-Book %

算法	Movie域						Book域
	k=5		k=10		k=20		k=5		k=10		k=20
	Recall	MRR	Recall	MRR	Recall	MRR	Recall	MRR	Recall	MRR	Recall	MRR
POP	0.18	0.07	0.29	0.11	0.58	0.13	0.20	0.11	0.44	0.14	0.75	0.16
Item-KNN	2.11	1.05	3.84	1.27	6.99	1.48	2.87	1.34	5.10	1.64	9.69	1.95
GRU4REC	13.69	12.79	14.11	12.70	14.43	12.88	14.64	13.79	15.02	13.82	15.34	13.95
NARM	14.20	13.80	14.53	13.85	14.80	13.96	15.57	15.25	15.66	15.26	15.78	15.27
STAMP	13.66	12.44	14.58	12.56	15.68	12.63	11.82	11.52	12.00	11.56	12.20	11.57
SR-GNN	12.66	11.77	14.18	13.54	14.87	13.88	15.43	15.12	15.61	15.14	15.77	15.15
CoNet	2.89	1.43	5.20	1.73	9.24	2.01	2.13	1.17	3.54	1.36	5.77	1.51
CDIE-C	9.56	7.31	10.45	8.96	12.18	9.45	9.30	7.18	10.11	7.25	11.29	7.49
CDHRM	13.05	12.60	13.97	12.55	14.12	12.74	14.89	13.91	15.33	14.27	15.50	14.32
π-Net	14.88	14.49	15.10	14.52	15.37	14.54	15.94	15.75	16.02	15.77	16.09	15.77
MIFN	15.13	14.84	16.34	15.07	16.56	15.38	16.57	16.05	16.66	16.16	16.73	16.23
TRaGCF	15.85	15.21	17.15	15.32	17.31	15.45	16.37	16.12	16.94	16.38	17.16	16.69

下载: 导出CSV

表 3 在Amazon Food-Kitchen上的实验结果

Table 3 Experimental Results on Amazon Food-Kitchen %

算法	Food域						Kitchen域
	k=5		k=10		k=20		k=5		k=10		k=20
	Recall	MRR	Recall	MRR	Recall	MRR	Recall	MRR	Recall	MRR	Recall	MRR
POP	0.83	0.40	1.50	0.49	2.15	0.56	0.39	0.22	0.70	0.25	1.47	0.27
Item-KNN	3.28	1.55	6.70	1.98	12.47	2.43	2.61	1.13	4.77	1.44	11.08	1.90
GRU4REC	9.46	8.10	10.38	8.23	11.26	8.29	8.70	8.36	8.93	8.39	9.25	8.41
NARM	10.34	9.43	11.86	9.54	12.23	9.62	8.69	8.41	8.91	8.44	9.21	8.46
STAMP	10.22	9.29	10.91	9.38	11.81	9.44	8.81	8.52	9.05	8.55	9.28	8.57
SR-GNN	10.68	9.31	11.02	9.49	11.67	9.60	9.12	8.31	9.36	8.62	9.57	8.92
CoNet	5.07	3.38	7.07	3.64	9.73	3.82	5.09	3.30	7.03	3.55	9.47	3.71
CDIE-C	7.98	6.44	9.36	7.12	10.75	7.79	8.57	8.34	8.95	8.39	9.38	8.45
CDHRM	9.56	8.21	10.40	9.66	11.49	9.35	8.62	8.41	9.07	8.43	9.41	8.50
π-Net	10.59	9.56	10.46	9.67	12.54	9.75	8.89	8.57	9.12	8.60	9.52	8.89
MIFN	11.20	9.91	12.25	10.16	13.27	10.25	9.72	9.18	10.01	9.21	10.33	9.23
TRaGCF	11.79	10.17	12.86	10.59	13.96	11.32	10.24	9.75	11.29	9.89	11.60	9.94

下载: 导出CSV

表 4 TRaGCF各功能模块贡献对比结果

Table 4 Comparison Results of the Contributions of Different Modules of TRaGCF %

算法	Food 域		Kitchen域
算法	Recall@20	MRR@20	Recall@20	MRR@20
TRaGCF-SBM	12.87	10.95	10.88	9.03
TRaGCF-CBM	13.42	11.03	11.41	9.79
TRaGCF	13.96	11.32	11.60	9.94

下载: 导出CSV

参考文献(38)

[1]	Fang Hui, Guo Guibing, Zhang Danning, et al. Deep learning-based sequential recommender systems: Concepts, algorithms, and evaluations[C] //Proc of the 19th Int Conf on Web Engineering. Berlin: Springer, 2019: 574−577
[2]	Wang Shoujin, Hu Liang, Wang Yan, et al. Sequential recommender systems: Challenges, progress and prospects[C] //Proc of the 28th Int Joint Conf Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2019: 6332−6338
[3]	Rendle S, Freudenthaler C, Schmidt-Thieme L. Factorizing personalized Markov chains for next-basket recommendation[C] //Proc of the 19th Int Conf on World Wide Web. New York: ACM, 2010: 811−820
[4]	Quadrana M, Karatzoglou A, Hidasi B, et al. Personalizing session-based recommendations with hierarchical recurrent neural networks[C] //Proc of the 11th ACM Conf on Recommender Systems. New York: ACM, 2017: 130−137
[5]	Tang Jiaxi, Wang Ke. Personalized top-n sequential recommendation via convolutional sequence embedding[C] //Proc of the 11th ACM Int Conf on Web Search and Data Mining. New York: ACM, 2018: 565−573
[6]	Wu Shu, Tang Yuyuan, Zhu Yanqiao, et al. Session-based recommendation with graph neural networks[C] //Proc of the 33rd AAAI Conf on Artificial Intelligence. Palo Alto, CA : AAAI, 2019: 346−353
[7]	Li Xiaohan, Zhang Mengqi, Wu Shu, et al. Dynamic graph collaborative filtering[C] //Proc of the 20th IEEE Int Conf on Data Mining (ICDM). Piscataway, NJ: IEEE, 2020: 322−331
[8]	Guo Yupu, Ling Yanxiang, Chen Honghui. A time-aware graph neural network for session-based recommendation[J]. IEEE Access, 2020, 8: 167371−167382
[9]	Qiu Ruihong, Li Jingjing, Huang Zi, et al. Rethinking the item order in session-based recommendation with graph neural networks[C] //Proc of the 28th ACM Int Conf on Information and Knowledge Management. New York: ACM, 2019: 579−588
[10]	王俊,李石君,杨莎,等. 一种新的用于跨领域推荐的迁移学习模型[J]. 计算机学报,2017,40(10):2367−2380 doi: 10.11897/SP.J.1016.2017.02367 Wang Jun, Li Shijun, Yang Sha, et al. A new transfer learning model for cross-domain recommendation[J]. Chinese Joural of Computers, 2017, 40(10): 2367−2380 (in Chinese) doi: 10.11897/SP.J.1016.2017.02367
[11]	Ma Muyuan, Ren Pengjie, Lin Yujie, et al. π-Net: A parallel information-sharing network for shared-account cross-domain sequential recommendations[C] //Proc of the 42nd Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2019: 685−694
[12]	郭磊,李秋菊,刘方爱,等. 基于自注意力网络的共享账户跨域序列推荐[J]. 计算机研究与发展,2021,58(11):2524−2537 doi: 10.7544/issn1000-1239.2021.20200564 Guo Lei, Li Qiuju, Liu Fang’ai, et al. Shared-account cross-domain sequential recommendation with self-attention network[J]. Journal of Computer Research and Development, 2021, 58(11): 2524−2537 (in Chinese) doi: 10.7544/issn1000-1239.2021.20200564
[13]	Ma Muyuan, Ren Pengjie, Chen Zhumin, et al. Mixed information flow for cross-domain sequential recommendations[J]. arXiv preprint, arXiv: 2012.00485, 2020
[14]	Yang Guang, Hong Xiaoguang, Peng Zhaohui, et al. Long short-term memory with sequence completion for cross-domain sequential recommendation[C] //Proc of the 4th Asia-Pacific Web (APWeb) and Web-Age Information Management (WAIM) Joint Int Conf on Web and Big Data. Berlin: Springer, 2020: 378−393
[15]	Wang Yaqin, Guo Caili, Chu Yunfei, et al. A cross-domain hierarchical recurrent model for personalized session-based recommendations[J]. Neurocomputing, 2020, 380(2): 271−284
[16]	Yuan Fajie, He Xiangnan, Karatzoglou A, et al. Parameter-efficient transfer from sequential behaviors for user modeling and recommendation[C] //Proc of the 43rd Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2020: 1469−1478
[17]	Kim D, Kim S, Zhao Handong, et al. Domain switch-aware holistic recurrent neural network for modeling multi-domain user behavior[C] //Proc of the 12th ACM Int Conf on Web Search and Data Mining. New York: ACM, 2019: 663−671
[18]	Wu Zonghan, Pan Shirui, Chen Fengwen, et al. A comprehensive survey on graph neural networks[J]. IEEE Transactions on Neural Networks and Learning Systems, 2020, 31(1): 4−24 doi: 10.1109/TNNLS.2019.2899262
[19]	Yu Juliang, Yin Hongzhi, Li Jundong, et al. Enhance social recommendation with adversarial graph convolutional networks[J]. arXiv preprint, arXiv: 2024.02340, 2020
[20]	Wang Wen, Zhang Wei, Liu Shukai, et al. Beyond clicks: Modeling multi-relational item graph for session-based target behavior prediction[C] //Proc of the 29th Web Conf 2020. New York: ACM, 2020: 3056−3062
[21]	Chen Tianwen, Wong R C-W. Handling information loss of graph neural networks for session-based recommendation[C] //Proc of the 26th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2020: 1172−1180
[22]	Xu Chengfeng, Zhao Pengpeng, Liu Yanchi, et al. Graph contextualized self-attention network for session-based recommendation[C] //Proc of the 28th Int Joint Conf on Artificial Intelligence. San Francisco, CA: Morgan Kaufmann, 2019: 3940−3946
[23]	Cui Zhihua, Xu Xianghua, Fei Xue, et al. Personalized recommendation system based on collaborative filtering for IoT scenarios[J]. IEEE Transactions on Services Computing, 2020, 13(4): 685−695 doi: 10.1109/TSC.2020.2964552
[24]	Wang Ziyang, Wei Wei, Cong Gao, et al. Global context enhanced graph neural networks for session-based recommendation[C] //Proc of the 43rd Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2020: 169−178
[25]	Ma Chen, Ma Liheng, Zhang Yingxue, et al. Memory augmented graph neural networks for sequential recommendation[C] //Proc of the 34th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2020, 34(3): 5045−5052
[26]	Sarwar B, Karypis G, Konstan J, et al. Item-based collaborative filtering recommendation algorithms[C] //Proc of the 10th Int Conf on World Wide Web. New York: ACM, 2001: 285−295
[27]	Wu Yuexin, Liu Hanxiao, Yang Yiming. Graph convolutional matrix completion for bipartite edge prediction[C] //Proc of the 10th Int Joint Conf on Knowledge Discovery, Knowledge Engineering and Knowledge Management. Berlin: Springer, 2018: 49−58
[28]	Wang Xiang, He Xiangnan, Wang Meng, et al. Neural graph collaborative filtering[C] //Pro of the 42nd Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2019: 165−174
[29]	Chen Lei, Wu Le, Hong Richang, et al. Revisiting graph based collaborative filtering: A linear residual graph convolutional network approach[C] //Proc of the 34th AAAI Conf on Artificial Intelligence. Palo Alto, CA: AAAI, 2020: 27−34
[30]	He Xiangnan, Deng Kuan, Wang Xiang, et al. LightGCN: Simplifying and powering graph convolution network for recommendation[C] //Proc of the 43rd Int ACM SIGIR Conf on Research and Development in Information Retrieval. New York: ACM, 2020: 639−648
[31]	Sun Jianing, Zhang Yingxue, Ma Chen, et al. Multi-graph convolution collaborative filtering[C] //Proc of the 19th IEEE Int Conf on Data Mining (ICDM). Piscataway, NJ: IEEE, 2019: 1306−1311
[32]	Schlichtkrull M, Kipf T N, Bloem P, et al. Modeling relational data with graph convolutional networks[C] //Proc of the 18th European Semantic Web Conf. Berlin: Springer, 2018: 593−607
[33]	Reddy S R S, Nalluri S, Kuniseti S, et al. Content-based movie recommendation system using genre correlation[M] //Smart Intelligent Computing and Applications. Berlin: Springer, 2019: 391−397
[34]	Hidasi B, Karatzoglou A, Baltrunas L, et al. Session-based recommendations with recurrent neural networks[J]. arXiv preprint, arXiv: 1511.06939, 2017
[35]	Li Jing, Ren Pengjie, Chen Zhumin, et al. Neural attentive session-based recommendation[C] //Proc of the 26th ACM on Conf on Information and Knowledge Management. New York: ACM, 2017: 1419−1428
[36]	Liu Qiao, Zeng Yifu, Mokhosi R, et al. STAMP: Short-term attention/memory priority model for session-based recommendation[C] //Proc of the 24th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining. New York: ACM, 2018: 1831−1839
[37]	Hu Guangneng, Zhang Yu, Yang Qiang. CoNet: Collaborative cross networks for cross-domain recommendation[C] //Proc of the 27th ACM Int Conf on Information and Knowledge Management. New York: ACM, 2018: 667−676
[38]	Wang Yaqing, Feng Chunyan, Guo Caili, et al. Solving the sparsity problem in recommendations via cross-domain item embedding based on co-clustering[C] //Proc of the 12th ACM Int Conf on Web Search and Data Mining. New York: ACM, 2019: 717−725

施引文献(3)

期刊类型引用(2)

1.	刘阳，鲁圆圆，郭成城. 基于优先级的数据中心任务优化调度算法设计. 计算机仿真. 2025(01): 497-500+507 . 百度学术
2.	骆海霞. 基于递推估计的Web前端偶发任务能耗感知方法. 黑龙江工业学院学报(综合版). 2023(10): 115-120 . 百度学术