基于知识图谱的轻量级图卷积网络推荐

doi:10.13232/j.cnki.jnju.2023.06.004

南京大学学报(自然科学版) ›› 2023, Vol. 59 ›› Issue (6): 937–946.doi: 10.13232/j.cnki.jnju.2023.06.004

基于知识图谱的轻量级图卷积网络推荐

彭永梅, 童向荣()

烟台大学计算机与控制工程学院，烟台，264005

收稿日期:2023-08-10 出版日期:2023-11-30 发布日期:2023-12-06
通讯作者: 童向荣 E-mail:xr_tong@163.com
基金资助:
国家自然科学基金(62072392);山东省重大科技创新工程(2019522Y020131);山东省自然科学基金(ZR2020QF113);烟台市重点实验室：高端海洋工程装备智能技术

Lightweight graph convolutional network recommendation based on knowledge graph

Yongmei Peng, Xiangrong Tong()

School of Computer and Control Engineering，Yantai University，Yantai，264005，China

Received:2023-08-10 Online:2023-11-30 Published:2023-12-06
Contact: Xiangrong Tong E-mail:xr_tong@163.com

摘要/Abstract

摘要：

知识图谱可以给推荐系统提供丰富的、结构化的信息，从而提高推荐准确性.最近的技术趋势是基于传播的方法设计端到端的模型，但有的基于传播的方法无法捕获项目的高阶协作信号.一般图卷积网络中包含的最常见形式是特征转换、非线性激活和邻域聚合，然而，经验表明，特征转换和非线性激活对协同过滤推荐不一定有积极的影响，更糟糕的是，它们可能会降低推荐性能，使训练更加困难.针对以上问题，提出基于知识图谱的轻量级图卷积网络推荐模型.首先，从实体邻居中抽取样本作为感受野，将知识图谱中的实体通过多次迭代嵌入传播来获取高阶邻域信息.感受野结合邻域信息和可能存在的偏差来计算实体表示，还可以扩展到多跳以模拟高阶连通性并捕获用户潜在的远距离兴趣.其次，使用邻域聚合以预测用户和项目之间的评分，这不仅简化了模型设计，还提高了模型的有效性和准确度.最后，在电影、书籍和音乐推荐这三个数据集中应用提出的模型，实验结果表明，提出的方法优于其他推荐基线.

关键词: 知识图谱, 推荐, 图卷积网络, 协同过滤, 嵌入传播

Abstract:

Knowledge graph provides the recommendation system with rich and structured information to improve the recommendation accuracy. Recent trend in technology is to design end?to?end models based on propagation，but some propagation?based approaches are unable to capture the higher?order collaboration signals of a project. The most common forms included in general graph convolutional networks are feature transformation，nonlinear activation and neighborhood aggregation. However，experience has shown that feature transformation and nonlinear activation do not necessarily have a positive effect on collaborative filtering recommendations，even worse，they may reduce recommendation performance and make training more difficult. To solve these problems，a lightweight graph convolutional network recommendation model based on knowledge graph is proposed. Firstly，samples from the physical neighbors are taken as receptive fields，and entities in the knowledge graph are embedded and propagated through multiple iterations to obtain higher?order neighborhood information. Receptive fields are combined with neighborhood information and possible deviations to calculate the entity representation. Receptive fields can be extended to multi?hops to simulate higher?order connectivity and capture users' potential long?distance interests. Secondly，neighborhood aggregation is used to predict ratings between users and projects，which not only simplifies model design，but also improves model validity and accuracy. Finally，the proposed model is applied to three datasets of movie，book and music recommendations，and experimental results show that the proposed method outperforms other recommendation baselines.

Key words: knowledge graph, recommendation, graph convolutional network, collaborative filtering, embedded propagation

中图分类号:

TP391

彭永梅, 童向荣. 基于知识图谱的轻量级图卷积网络推荐[J]. 南京大学学报(自然科学版), 2023, 59(6): 937–946.

Yongmei Peng, Xiangrong Tong. Lightweight graph convolutional network recommendation based on knowledge graph[J]. Journal of Nanjing University(Natural Sciences), 2023, 59(6): 937–946.

图/表 7

图1

表1

表2

表3

表4

表5

图2

参考文献 28

1	He X N， He Z K， Song J K，et al. NAIS：Neural attentive item similarity model for recommendation. IEEE Transactions on Knowledge and Data Engineering，2018，30(12)：2354-2366.
2	Wang H W， Zhang F Z， Xie X，et al. DKN：Deep knowledge?aware network for news recommen?dation∥Proceedings of 2018 World Wide Web Conference. Lyon，France：ACM，2018：1835-1844.
3	Wang H W， Zhang F Z， Wang J L，et al. RippleNet：Propagating user preferences on the knowledge graph for recommender systems∥Proceedings of the 27th ACM International Conference on Information and Knowledge Management. Torino，Italy：ACM，2018：417-426.
4	Hamilton W L， Ying R， Leskovec J. Inductive representation learning on large graphs∥Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach，CA，USA：ACM，2017：1025-1035.
5	Wang X， He X N， Wang M，et al. Neural graph collaborative filtering∥Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. Paris，France：ACM，2019：165-174.
6	He X N， Deng K， Wang X，et al. LightGCN：Simplifying and powering graph convolution network for recommendation∥Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. Virtual Event：ACM，2020：639-648.
7	张明星，张骁雄，刘姗姗，等. 利用知识图谱的推荐系统研究综述. 计算机工程与应用，2023，59(4)：30-42.
	Zhang M X， Zhang X X， Liu S S，et al. Review of recommendation systems using knowledge graph. Computer Engineering and Applications，2023，59(4)：30-42.
8	Bordes A， Usunier N， Garcia?Durán A，et al. Translating embeddings for modeling multi?relational data∥Proceedings of the 26th International Conference on Neural Information Processing Systems. Lake Tahoe，Nevada：ACM，2013：2787-2795.
9	Wang Z， Zhang J W， Feng J L，et al. Knowledge graph embedding by translating on hyperplanes∥Proceedings of the 28th AAAI Conference on Artificial Intelligence. Québec City，Canada：AAAI Press，2014：1112-1119.
10	Chen Z X， Wang X T， Xie X，et al. Towards explainable conversational recommendation∥Proceedings of the 29th International Joint Conference on Artificial Intelligence. Yokohama，Japan：ACM，Article No.414，2021.
11	Wang H W， Zhang F Z， Zhao M，et al. Multi?task feature learning for knowledge graph enhanced recommendation∥The World Wide Web Conference. San Francisco，CA，USA：ACM，2019：2000-2010.
12	Cao Y X， Wang X， He X N，et al. Unifying knowledge graph learning and recommendation：Towards a better understanding of user preferences∥The World Wide Web Conference. San Francisco，CA，USA：ACM，2019：151-161.
13	罗承天，叶霞. 基于知识图谱的推荐算法研究综述. 计算机工程与应用，2023，59(1)：49-60.
	Luo C T， Ye X. Survey on knowledge graph?based recommendation methods. Computer Engineering and Appli?cations，2023，59(1)：49-60.
14	Xian Y K， Fu Z H， Muthukrishnan S，et al. Reinforcement knowledge graph reasoning for explainable recommendation∥Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. Paris，France：ACM，2019：285-294.
15	Wang X， Huang T L， Wang D X，et al. Learning intents behind interactions with knowledge graph for recommendation∥Proceedings of the Web Conference 2021. Ljubljana，Slovenia：ACM，2021：878-887.
16	Wang H， Zhao M， Xie X，et al. Knowledge graph convolutional networks for recommender systems∥Proceedings of the 2019 World Wide Web Conference. San Francisco，CA，USA：ACM，2019：3307-3313.
17	Ai Q Y， Azizi V， Chen X，et al. Learning heterogeneous knowledge base Embeddings for explainable recommendation. Algorithms，2018，11(9)：137.
18	葛尧，陈松灿. 面向推荐系统的图卷积网络. 软件学报，2020，31(4)：1101-1112.
	Ge Y， Chen S C. Graph convolutional network for recommender systems. Journal of Software，2020，31(4)：1101-1112.
19	Bruna J， Zaremba W， Szlam A，et al. Spectral networks and locally connected networks on graphs∥The 2nd International Conference on Learning Representations. Banff，Canada：ICLR，2014，DOI:10.48550/arXiv.1312.6203 .
20	Defferrard M， Bresson X， Vandergheynst P. Convolutional neural networks on graphs with fast localized spectral filtering∥Proceedings of the 30th International Conference on Neural Information Processing Systems. Barcelona，Spain：ACM，2016：3844-3852.
21	Kipf T N， Welling M. Semi?supervised classification with graph convolutional networks∥The 5th International Conference on Learning Represen?tations. Toulon，France：ICLR，2017，DOI:10.48550/arXiv.1312.6203 .
22	Ying R， He R N， Chen K F，et al. Graph convolutional neural networks for web?scale recom?mender systems∥Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. London，UK：ACM，2018：974-983.
23	Wang X， Jin H Y， Zhang A，et al. Disentangled graph collaborative filtering∥Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval. Virtual Event：ACM，2020：1001-1010.
24	Koren Y. Factorization meets the neighborhood：A multifaceted collaborative filtering model∥Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Las Vegas，NV，USA：ACM，2008：426-434.
25	Rendle S. Factorization machines with libfm. ACM Transactions on Intelligent Systems and Technology，2012，3(3)：57.
26	Yu X， Ren X， Sun Y Z，et al. Personalized entity recommendation：A heterogeneous information network approach∥Proceedings of the 7th ACM International Conference on Web Search and Data Mining. New York，NY，USA：ACM，2014：283-292.
27	Zhang FZ， Yuan NJ， Lian DF，et al. Collaborative knowledge base embedding for recommender systems∥Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco，CA，USA：ACM，2016：353-362.
28	Wang X， He X N， Cao Y X，et al. KGAT：Knowledge graph attention network for recommen?dation∥Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. Anchorage，AK，USA：ACM，2019：950-958，DOI：10.1145/3292500.3330989 .

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

	MovieLens⁃20M	Book⁃Crossing	Last.FM
#users	138159	19676	1872
#items	16954	20003	3846
#interactions	13501622	172576	42346
#entities	102569	25787	9366
#relations	32	18	60
#KG triples	499474	60787	15518
K	4	8	8
d	32	64	16
H	2	1	1
λ	10^-7	2×10^-5	10^-4
η	2×10^-2	2×10^-4	5×10^-4
batch size	65536	256	128

K	2	4	8	16	32	64
MovieLens⁃20M	0.978	0.980	0.978	0.978	0.976	0.977
Book⁃Crossing	0.680	0.723	0.739	0.724	0.720	0.730
Last.FM	0.790	0.795	0.797	0.793	0.795	0.792

H	1	2	3	4
MovieLens⁃20M	0.976	0.980	0.973	0.625
Book⁃Crossing	0.749	0.743	0.675	0.547
Last.FM	0.798	0.739	0.564	0.536

d	4	8	16	32	64
MovieLens⁃20M	0.973	0.977	0.979	0.975	0.973
Book⁃Crossing	0.735	0.737	0.739	0.741	0.738
Last.FM	0.795	0.798	0.793	0.791	0.789

Model	MovieLens⁃20M		Book⁃Crossing		Last.FM
Model	AUC	F1	AUC	F1	AUC	F1
LightKGCN	0.980	0.936	0.747	0.701	0.810	0.735
SVD	0.963	0.919	0.612	0.635	0.769	0.696
LibFM	0.959	0.906	0.691	0.618	0.778	0.710
LibFM+TransE	0.966	0.917	0.698	0.622	0.777	0.709
PER	0.832	0.788	0.617	0.562	0.633	0.596
CKE	0.924	0.871	0.677	0.611	0.744	0.673
RippleNet	0.950	0.912	0.715	0.650	0.776	0.702
KGAT	0.965	0.926	0.735	0.681	0.788	0.722
KGCN	0.977	0.930	0.738	0.688	0.776	0.708

基于知识图谱的轻量级图卷积网络推荐

Lightweight graph convolutional network recommendation based on knowledge graph

RichHTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 28

相关文章 15

Metrics

本文评价

推荐阅读 0

[1]	黄树添, 胡诗琳, 卜祥智, 李华雄, 刘久兵. 融入用户风险偏好的三支协同过滤推荐模型[J]. 南京大学学报(自然科学版), 2023, 59(5): 777-789.
[2]	边纪超, 庞继芳, 宋鹏. 融合隐式信任与属性偏好的群组推荐算法[J]. 南京大学学报(自然科学版), 2023, 59(5): 803-812.
[3]	武桐舟, 刘强, 王亮. 基于不确定性估计的推荐系统多任务学习去偏方法[J]. 南京大学学报(自然科学版), 2023, 59(4): 543-549.
[4]	刘思源, 毛存礼, 张勇丙. 基于领域知识图谱和对比学习的汉越跨境民族文本检索方法[J]. 南京大学学报(自然科学版), 2023, 59(4): 610-619.
[5]	刘志中, 李林霞, 孟令强. 基于混合图神经网络的个性化POI推荐方法研究[J]. 南京大学学报(自然科学版), 2023, 59(3): 373-387.
[6]	冯海, 马甲林, 许林杰, 杨宇, 谢乾. 融合标签嵌入和知识感知的多标签文本分类方法[J]. 南京大学学报(自然科学版), 2023, 59(2): 273-281.
[7]	刘佳慧, 袁卫华, 曹家伟, 张涛, 张志军. 基于用户特征聚类与服务质量预测的推荐方法[J]. 南京大学学报(自然科学版), 2023, 59(1): 120-133.
[8]	韩迪, 陈怡君, 廖凯, 林坤玲. 推荐系统中的准确性、新颖性和多样性的有效耦合与应用[J]. 南京大学学报(自然科学版), 2022, 58(4): 604-614.
[9]	吕亚兰, 徐媛媛, 张恒汝. 一种可解释性泛化矩阵分解推荐算法[J]. 南京大学学报(自然科学版), 2022, 58(1): 135-142.
[10]	武聪, 马文明, 王冰, 朱建豪. 融合用户标签相似度的矩阵分解算法[J]. 南京大学学报(自然科学版), 2022, 58(1): 143-152.
[11]	黄治纲, 谢新强, 邢铁军, 葛东, 蔡晨秋, 窦丽莉, 王天翊. 基于司法案例知识图谱的类案推荐[J]. 南京大学学报(自然科学版), 2021, 57(6): 1053-1063.
[12]	郝昱猛, 马文明, 王冰. 基于特定用户约束的概率矩阵分解算法[J]. 南京大学学报(自然科学版), 2021, 57(5): 818-827.
[13]	袁晓峰, 钱苏斌, 周彩根. 基于填充先验约束的矩阵分解算法[J]. 南京大学学报(自然科学版), 2021, 57(2): 197-207.
[14]	张家精,夏巽鹏,陈金兰,倪友聪. 基于张量分解和深度学习的混合推荐算法[J]. 南京大学学报(自然科学版), 2019, 55(6): 952-959.
[15]	徐媛媛,张恒汝,闵帆,黄雨婷. 三支交互推荐[J]. 南京大学学报(自然科学版), 2019, 55(6): 973-983.