基于元路径卷积的异构图神经网络算法

doi:10.11959/j.issn.1000-0801.2024070

Abstract

Abstract:

In the multilayer graph convolution calculation, each node is usually represented as a single vector, which makes the high-order graph convolution layer unable to distinguish the information of different relationships and sequences, resulting in the loss of information in the transmission process.To solve this problem, a heterogeneous graph neural network algorithm based on meta-path convolution was proposed.Firstly, the feature transformation was used to adaptively adjust the node features.Secondly, the high-order indirect relationship between the nodes was mined by convolution within the meta-path to capture the interaction between the target node and other types of nodes under the element path.Finally, the reciprocity between semantics was explored through the self-attention mechanism, and the features from different meta-paths were fused.Extensive experiments were carried out on ACM, IMDB and DBLP datasets, and compared with the current mainstream algorithms.The experimental results show that the average increase of Macro-F1 in the node classification task is 0.5%～3.5%, and the ARI value in the node clustering task is increased by 1%～3%, which proves that the algorithm is effective and feasible.

Key words: heterogeneous graph, graph embedding, graph neural network, meta-path, graph convolution

CLC Number:

TP391

Zhilong QIN, Kun DENG, Xingyan LIU. Meta-path convolution based heterogeneous graph neural network algorithm[J]. Telecommunications Science, 2024, 40(3): 89-103.

Figures/Tables 11

	指标	训练集比例	metapath2vec	GCN	HAN	MAGNN	HGNN_AC	HGB	OSGNN	MCHGNN
IMDB	Macro-F1	20%	46.3%	45.0%	57.4%	57.9%	59.7%	60.7%	59.5%	$61 . 2 %$
		40%	47.7%	45.4%	57.5%	59.1%	60.2%	61.2%	60.4%	$63 . 8 %$
		60%	48.2%	46.8%	57.7%	59.6%	60.6%	62.6%	60.8%	$65 . 1 %$
		80%	48.9%	47.2%	57.8%	59.8%	60.8%	63.2%	61.0%	$66 . 6 %$
	Micro-F1	20%	48.2%	46.3%	57.2%	58.0%	59.8%	60.8%	59.5%	$61 . 2 %$
		40%	49.5%	48.1%	57.6%	59.3%	60.4%	61.4%	60.4%	$63 . 8 %$
		60%	50.2%	48.5%	57.7%	59.7%	60.8%	62.8%	60.8%	$65 . 0 %$
		80%	50.8%	48.6%	57.8%	60.0%	61.0%	63.2%	61.0%	$66 . 6 %$
ACM	Macro-F1	20%	70.0%	70.2%	90.1%	88.9%	89.2%	91.1%	$92 . 6 %$	92.0%
		40%	71.3%	70.7%	90.2%	89.7%	90.7%	92.4%	$93 . 3 %$	92.2%
		60%	71.8%	70.6%	91.4%	90.2%	91.5%	92.6%	$93 . 7 %$	93.0%
		80%	72.3%	68.4%	91.5%	90.3%	91.9%	92.9%	93.8%	$94 . 0 %$
	Micro-F1	20%	72.3%	73.2%	89.9%	88.7%	89.1%	90.2%	$92 . 6 %$	92.2%
		40%	73.5%	73.6%	90.3%	89.2%	90.7%	91.3%	$93 . 1 %$	92.3%
		60%	73.9%	74.2%	91.4%	90.3%	91.5%	91.5%	93.6%	$93 . 7 %$
		80%	74.2%	73.9%	91.7%	90.8%	91.9%	92.3%	93.7%	$93 . 8 %$
DBLP	Macro-F1	20%	91.6%	90.1%	91.6%	92.4%	94.2%	94.0%	94.0%	$94 . 8 %$
		40%	92.8%	90.2%	91.6%	93.5%	94.4%	94.1%	94.3%	$95 . 0 %$
		60%	93.0%	90.4%	92.1%	93.9%	94.4%	94.2%	94.5%	$94 . 8 %$
		80%	93.2%	90.7%	92.2%	93.9%	94.6%	94.3%	94.8%	$95 . 3 %$
	Micro-F1	20%	92.2%	90.4%	92.3%	92.9%	94.6%	94.2%	94.4%	$95 . 3 %$
		40%	93.2%	90.9%	92.5%	93.0%	94.7%	94.5%	94.7%	$95 . 4 %$
		60%	93.9%	91.0%	92.6%	93.4%	94.8%	94.5%	94.9%	$95 . 2 %$
		80%	94.0%	91.1%	92.8%	93.9%	95.0%	94.6%	95.2%	$95 . 6 %$

算法	IMDB		ACM		DBLP
算法	NMI	ARI	NMI	ARI	NMI	ARI
MAGNN	13.3%	11.6%	65.5%	69.3%	79.2%	84.6%
HGNN_AC	13.6%	14.4%	64.9%	68.5%	79.6%	84.9%
OSGNN	15.9%	18.1%	72.4%	74.5%	81.6%	86.4%
MCHGNN	$16 . 8 %$	$19 . 4 %$	$74 . 5 %$	$77 . 1 %$	$83 . 1 %$	$87 . 4 %$

References 29

[1]	WEI Y C , FU X C , SUN Q Y ,et al. Heterogeneous graph neural network for privacy-preserving recommendation[C]// Proceedings of the 2022 IEEE International Conference on Data Mining (ICDM). Piscataway:IEEE Press, 2022: 528-537.
[2]	杨帅, 王瑞琴, 马辉 . 基于多通道的边学习图卷积网络[J]. 电信科学, 2022,38(9): 95-104.
	YANG S , WANG R Q , MA H . Multi-channel based edgelearning graph convolutional network[J]. Telecommunications Science, 2022,38(9): 95-104.
[3]	WANG J C , ZHANG Y , WANG L X ,et al. Multitask hypergraph convolutional networks:a heterogeneous traffic prediction framework[J]. IEEE Transactions on Intelligent Transportation Systems, 2022,23(10): 18557-18567.
[4]	刘群, 谭洪胜, 张优敏 ,等. 基于元路径的动态异质网络表示学习[J]. 电子学报, 2022,50(8): 1830-1839.
	LIU Q , TAN H S , ZHANG Y M ,et al. Dynamic heterogeneous network representation method based on meta-path[J]. Acta Electronica Sinica, 2022,50(8): 1830-1839.
[5]	YU P Y , FU C F , YU Y W ,et al. Multiplex heterogeneous graph convolutional network[C]// Proceedings of the Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York:ACM, 2022: 2377-2387.
[6]	陶涛, 李珍, 王冀彬 ,等. 基于图神经网络的权益推荐技术方案研究[J]. 电信科学, 2023,39(8): 91-101.
	TAO T , LI Z , WANG J B ,et al. Research on the graphical convolution neural network based benefits recommendation system strategy[J]. Telecommunications Science, 2023,39(8): 91-101.
[7]	KIPF T N , WELLING M . Semi-supervised classification with graph convolutional networks[J]. arXiv preprint,arXiv:1609.02907, 2016.
[8]	VELIKOVI P , CUCURULL G , CASANOVA A ,et al. Graph attention networks[J]. arXiv preprint,arXiv:1710.10903, 2017.
[9]	WANG X , JI H Y , SHI C ,et al. Heterogeneous graph attention network[C]// Proceedings of the WWW '19:The World Wide Web Conference. New York:ACM, 2019: 2022-2032.
[10]	FU X Y , ZHANG J N , MENG Z Q ,et al. MAGNN:metapath aggregated graph neural network for heterogeneous graph embedding[C]// Proceedings of the Proceedings of The Web Conference 2020. New York:ACM, 2020: 2331-2341.
[11]	YANG X C , YAN M Y , PAN S R ,et al. Simple and efficient heterogeneous graph neural network[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2023,37(9): 10816-10824.
[12]	ZHAO T Y , YANG C , LI Y B ,et al. Space4HGNN:a novel,modularized and reproducible platform to evaluate heterogeneous graph neural network[C]// Proceedings of the Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York:ACM, 2022: 2776-2789.
[13]	任嘉睿, 张海燕, 朱梦涵 ,等. 基于元图卷积的异质网络嵌入学习算法[J]. 计算机研究与发展, 2022,59(8): 1683-1693.
	REN J R , ZHANG H Y , ZHU M H ,et al. Embedding learning algorithm for heterogeneous network based on meta-graph convolution[J]. Journal of Computer Research and Development, 2022,59(8): 1683-1693.
[14]	JIN D , HUO C Y , LIANG C D ,et al. Heterogeneous graph neural network via attribute completion[C]// Proceedings of the Proceedings of the Web Conference 2021. New York:ACM, 2021: 391-400.
[15]	XU W T , XIA Y C , LIU W Q ,et al. Shgnn:structure-aware heterogeneous graph neural network[J]. arXiv preprint,arXiv:2112.06244, 2021.
[16]	ZHAN L , JIA T . CoarSAS2hvec:heterogeneous information network embedding with balanced network sampling[J]. Entropy, 2022,24(2): 276.
[17]	YUN S , JEONG M , KIM R ,et al. Graph transformer networks[J]. Advances in neural information processing systems, 2019,32.
[18]	YANG Y M , GUAN Z Y , LI J X ,et al. Interpretable and efficient heterogeneous graph convolutional network[J]. IEEE Transactions on Knowledge and Data Engineering, 2023,35(2): 1637-1650.
[19]	CHEN K J , LU H , LIU Z ,et al. Heterogeneous graph convolutional network with local influence[J]. Knowledge-Based Systems, 2022(236): 107699.
[20]	LYU Q S , DING M , LIU Q ,et al. Are we really making much progress? :Revisiting,benchmarking and refining heterogeneous graph neural networks[C]// Proceedings of the Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery ＆ Data Mining. New York:ACM, 2021: 1150-1160.
[21]	DU C G , YAO K C , ZHU H S ,et al. Seq-HGNN:learning sequential node representation on heterogeneous graph[J]. arXiv preprint,arXiv:2305.10771, 2023.
[22]	王悦天, 傅司超, 彭勤牧 ,等. 半监督场景下多视角信息交互的图卷积神经网络[J]. 软件学报, 2023: 1-18.
	WANG Y T , FU S C , PENG Q M ,et al. Multi-view interaction graph convolutional network for semi-supervised classification[J]. Journal of Software, 2023: 1-18.
[23]	WANG X , BO D Y , SHI C ,et al. A survey on heterogeneous graph embedding:methods,techniques,applications and sources[J]. IEEE Transactions on Big Data, 2023,9(2): 415-436.
[24]	GUAN W L , JIAO F K , SONG X M ,et al. Personalized fashion compatibility modeling via metapath-guided heterogeneous graph learning[C]// Proceedings of the Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York:ACM, 2022: 482-491.
[25]	WU Z Y , LIANG Q Y , ZHAN Z H . Course recommendation based on enhancement of meta-path embedding in heterogeneous graph[J]. Applied Sciences, 2023,13(4): 2404.
[26]	CHEN H , HONG P F , HAN W ,et al. Dialogue relation extraction with document-level heterogeneous graph attention networks[J]. Cognitive Computation, 2023,15(2): 793-802.
[27]	DONG Y X , CHAWLA N V , SWAMI A . metapath2vec:scalable representation learning for heterogeneous networks[C]// Proceedings of the Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York:ACM, 2017: 135-144.
[28]	YAN Y Y , LI C , YU Y W ,et al. OSGNN:original graph and subgraph aggregated graph neural network[J]. Expert Systems with Applications, 2023(225): 120115.
[29]	VAN DER M L , HINTON G . Visualizing data using t-SNE[J]. Journal of Machine Learning Research, 2008(9): 2579-2605.

Metrics

Recommended 0

No Suggested Reading articles found!

数据集	节点类型	边类型	元路径
IMDB	M：4 287	M-D：4 278	M-D-M
	D：2 081	M-A：12 828	M-A-M
	A：5 257
ACM	P：4 287	P-P：9 615	P-A-P
	A：7 167	P-A：13 407	P-S-P
	S：60	P-S：4 019
DBLP	A：4 057	A-P：19 645	A-P-A
	P：14 328	P-T：85 810	A-P-T-P-A
	T：7 723	P-V：14 328	A-P-V-P-A
	V：20

元路径	节点类型	重要性权值
MDM	M	0.11×0.61=0.067 1
	MD	0.37×0.61=0.225 7
	MDM	0.52×0.61=0.317 2
MAM	M	0.20×0.39=0.078 0
	MA	0.42×0.39=0.163 8
	MAM	0.38×0.39=0.148 2

Meta-path convolution based heterogeneous graph neural network algorithm

RichHTML

PDF下载

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 11

References 29

Related Articles 6

Metrics

Recommended 0

[1]	Ruiqin WANG, Yimin HUANG, Qishun JI, Chaoyi WAN, Zhifeng ZHOU. Attention aware edge-node exchange graph neural network [J]. Telecommunications Science, 2024, 40(1): 106-114.
[2]	Xiongtao ZHANG, Jingyu ZHENG, Qing SHEN, Danfeng SUN, Yunliang JIANG. Multi-channel spatial-temporal traffic flow prediction based on hybrid static-dynamic graph convolution [J]. Telecommunications Science, 2023, 39(9): 97-110.
[3]	Tao TAO, Zhen LI, Jibin WANG, Haiyong XU, Yong JIANG, Zhuo CEHN, Runbo ZHANG, Qingyuan HU. Research on the graphical convolution neural network based benefits recommendation system strategy [J]. Telecommunications Science, 2023, 39(8): 91-101.
[4]	Shuai YANG, Ruiqin WANG, Hui MA. Multi-channel based edge-learning graph convolutional network [J]. Telecommunications Science, 2022, 38(9): 95-104.
[5]	Yijiang LI, Huibiao YE, Renhua XIE, Jiali LOU, Danna ZHUANG, Chuanhuang LI. Intelligent prediction method of network performance based on graph neural network [J]. Telecommunications Science, 2022, 38(3): 143-157.
[6]	Shaoqing WU,Yihong DONG,Xiong WANG,Yan CAO,Yu XIN. Learning attribute network algorithm based on high-order similarity [J]. Telecommunications Science, 2020, 36(12): 20-32.