一种融合局部拓扑影响力的时序链路预测算法

朱宇航; 刘树新; 吉立新; 何赞园; 李英乐

doi:10.11999/JEIT210019

一种融合局部拓扑影响力的时序链路预测算法

doi: 10.11999/JEIT210019

中国人民解放军战略支援部队信息工程大学信息技术研究所郑州 450002

基金项目: 国家自然科学基金(61521003, 61803384)

详细信息

作者简介:
朱宇航：男，1982年生，副教授，研究方向为链路预测、网络行为分析

刘树新：男，1987年生，助理研究员，研究方向为复杂网络演化、链路预测、通信网络安全

吉立新：男，1969年生，研究员，研究方向为电信网安全、社团发现

何赞园：男，1975年生，副研究员，研究方向为计算机通信、网络安全

李英乐：男，1985年生，副研究员，研究方向为社会网络分析、通信网络安全

通讯作者:
刘树新　liushuxin11@126.com

中图分类号: TN915; TP391
计量
- 文章访问数: 810
- HTML全文浏览量: 603
- PDF下载量: 110
- 被引次数: 0
出版历程
- 收稿日期: 2021-01-06
- 修回日期: 2021-08-23
- 网络出版日期: 2021-09-13
- 刊出日期: 2022-04-18

A Temporal Link Predict Algorithm Based on Fusion Local Structure Influence

Institute of Information Technology, PLA Strategic Support Force Information Engineering University, Zhengzhou 450002, China

Funds: The National Natural Science Foundation of China (61521003, 61803384)

摘要

摘要: 链路预测旨在发现复杂网络中的未知连接和未来可能的连接，在推荐系统等实际应用中具有重要作用。考虑到许多真实网络的时序特性，时序链路预测逐渐成为研究热点。当前，基于时间序列分析的方法往往忽略了网络演化过程对网络本身的影响，而基于静态网络演化的方法大多仅考虑了局部连边的演化影响，对网络拓扑结构的演化特性挖掘有限。针对上述问题，该文提出一种融合局部拓扑影响力的时序链路预测算法(TLP-FLSI)。首先，基于网络拓扑结构影响力作用，提出时序链路预测的通用模型(CTLPM)；其次，研究拓扑实体间相互作用在动态网络上的演化规律，分别定义了节点和连边的演化因子，以及时间序列衰减的演化因子，综合利用多个维度的特征信息，给出了融合局部节点和连边特征影响力的时序链路预测算法；最后，在7个真实数据集上分别进行实验，对比传统基于移动平均方法、误差修正、邻居扩展加权和图注意力网络等时序链路预测方法，实验结果证明该算法具有较好的准确率和排序性能。
- 时序动态网络 /
- 链路预测 /
- 融合特征 /
- 相似性指标 /
- 影响力衰减
Abstract: Link prediction aims to discover missing connected edges and possible future interaction in complex networks. The evolution mechanism of temporal networks has gained the attention of researchers with its ubiquitous applications in a variety of real-world scenarios. At present, many methods based on time series analysis are proposed, but the influence of the network evolution process on the network itself is ignored, and the methods based on the static network algorithm only consider the influence of the evolution of edges, which may lead to inadequate utilization of feature information and can not achieve better prediction accuracy. In view of the above problems, a novel Temporal Link Prediction algorithm base on Fusion Local Structure Influence (TLP-FLSI) is proposed, which fuses the impact of local nodes and edges. Firstly, based on the influence of network topology structure, Common Temporal Link Prediction Model(CTLPM)is proposed. Secondly, the evolution mechanism of the interaction between topological entities on the dynamic network is studied, and the evolution factors of nodes and edges, as well as the decay evolution factors of time series are defined respectively, and considering various factors, TLP-FLSI is derivated from CTLPM. Finally, compared with traditional temporal link predict method, including moving average methods, error correction methods, extended weighted method, graph attention methods, experimental results of seven real data sets show that TLP-FLSI achieves great improvement in accuracy and ranking score.
- Temporal dynamic network /
- Link prediction /
- Fusion features /
- Similarity metrics /
- Influence decay

HTML全文

图 1 复杂网络示意图

下载: 全尺寸图片幻灯片

图 2 TLP-FLSI与CTLPM的关系图

下载: 全尺寸图片幻灯片

图 3 节点局部影响力演化示意图

下载: 全尺寸图片幻灯片

图 4 连边局部影响力演化示意图

下载: 全尺寸图片幻灯片

图 5 影响力时序演化示意图

下载: 全尺寸图片幻灯片

图 6 不同方法的平均RS性能对比图

下载: 全尺寸图片幻灯片

图 7 步骤1实验权重参数影响的变化曲线

下载: 全尺寸图片幻灯片

8 步骤2实验权重参数影响的变化曲线

下载: 全尺寸图片幻灯片

图 9 步骤3实验权重参数影响的变化曲线

下载: 全尺寸图片幻灯片

表 1 数据集参数表

	Email	Enron	Facebook	DNC	MAN	UCI	LEM
节点数	1005	87273	63891	2029	167	1899	485
连边数	332334	1048576	855542	39264	82927	59835	196364
快照图数	76	66	52	33	39	28	51
时序周期	周	周	2周	天	周	周	半年

下载: 导出CSV

表 2 不同方法的平均AUC性能结果

模型	Email	Enron	Facebook	DNC	MAN	UCI	LEM
Last	0.7535	0.6056	0.5158	0.7011	0.8415	0.5152	0.9021
AV	0.8793	0.7282	0.5350	0.8156	0.9064	0.5724	0.9620
TS-W	0.8993	0.7646	0.5422	0.8357	0.9099	0.6170	0.9682
TLP-I	0.9051	0.7480	0.5526	0.8407	0.8989	0.5864	0.9692
LPE	0.8202	0.7728	0.6507	0.8264	0.8901	0.6699	0.8623
Reduce	0.8954	0.7732	0.5522	0.8252	0.8925	0.6157	0.9630
TMDN	0.8447	0.8103	0.6745	0.8561	0.8939	0.6995	0.8631
DySAT	0.8603	0.7885	0.7515	0.8376	0.9020	0.7187	0.8352
TASM	0.8501	0.7945	0.7702	0.8462	0.8975	0.7230	0.8373
FLSI-I(1)	0.9101	0.8551	0.6860	0.9148	0.9199	0.7192	0.9637
FLSI-I(2)	0.9092	0.7753	0.6193	0.8907	0.9306	0.6137	0.9669
FLSI-II	0.9246	0.8803	0.6973	0.9347	0.9264	0.7574	0.9709
FLSI-III	0.9111	0.8551	0.6860	0.9152	0.9180	0.7220	0.9650

下载: 导出CSV

参考文献(32)

[1]	PORTER M A. Nonlinearity + Networks: A 2020 Vision[M]. KEVREKIDIS P G, CUEVAS-MARAVER J, and SAXENA A. Emerging Frontiers in Nonlinear Science. Cham: Springer, 2020: 131–159. doi: 10.1007/978-3-030-44992-6_6.
[2]	WANG Yongcheng, WANG Yu, LIN Xinye, et al. The influence of network structural preference on link prediction[J]. Discrete Dynamics in Nature and Society, 2020, 2020: 6148273. doi: 10.1155/2020/6148273
[3]	KERRACHE S, ALHARBI R, and BENHIDOUR H. A scalable similarity-popularity link prediction method[J]. Scientific Reports, 2020, 10(1): 6394. doi: 10.1038/s41598-020-62636-1
[4]	DIVAKARAN A and MOHAN A. Temporal link prediction: A survey[J]. New Generation Computing, 2020, 38(1): 213–258. doi: 10.1007/s00354-019-00065-z
[5]	李治成, 吉立新, 刘树新, 等. 基于拓扑稳定性的有向网络链路预测方法[J]. 计算机应用研究, 2020, 37(12): 3744–3748. doi: 10.19734/j.issn.1001-3695.2019.09.0550 LI Zhicheng, JI Lixin, LIU Shuxin, et al. Link prediction method based on topological stability in directed network[J]. Application Research of Computers, 2020, 37(12): 3744–3748. doi: 10.19734/j.issn.1001-3695.2019.09.0550
[6]	王凯, 刘树新, 陈鸿昶, 等. 一种基于节点间资源承载度的链路预测方法[J]. 电子与信息学报, 2019, 41(5): 1225–1234. doi: 10.11999/JEIT180553 WANG Kai, LIU Shuxin, CHEN Hongchang, et al. A new link prediction method for complex networks based on resources carrying capacity between nodes[J]. Journal of Electronics &Information Technology, 2019, 41(5): 1225–1234. doi: 10.11999/JEIT180553
[7]	刘树新, 李星, 陈鸿昶, 等. 基于资源传输匹配度的复杂网络链路预测方法[J]. 通信学报, 2020, 41(6): 70–79. doi: 10.11959/j.issn.1000-436x.2020124 LIU Shuxin, LI Xing, CHEN Hongchang, et al. Link prediction method based on matching degree of resource transmission for complex network[J]. Journal on Communications, 2020, 41(6): 70–79. doi: 10.11959/j.issn.1000-436x.2020124
[8]	DA SILVA SOARES P R and PRUDÊNCIO R B C. Time series based link prediction[C]. The 2012 International Joint Conference on Neural Networks (IJCNN), Brisbane, Australia, 2012, 20: 1–7.
[9]	HUANG Zan and LIN D K J. The time-series link prediction problem with applications in communication surveillance[J]. INFORMS Journal on Computing, 2009, 21(2): 286–303. doi: 10.1287/ijoc.1080.0292
[10]	GÜNEŞ İ, GÜNDÜZ-ÖĞÜDÜCÜ Ş, and ÇATALTEPE Z. Link prediction using time series of neighborhood-based node similarity scores[J]. Data Mining and Knowledge Discovery, 2016, 30(1): 147–180. doi: 10.1007/s10618-015-0407-0
[11]	刘继嘉, 王童, 何兴盛, 等. 一种基于混合结构的动态网络链路预测算法[J]. 电子技术, 2018, 47(7): 53–59. doi: 10.3969/j.issn.1000-0755.2018.07.015 LIU Jijia, WANG Tong, HE Xingsheng, et al. A link prediction method of dynamic network based on hybrid structure[J]. Electronic Technology, 2018, 47(7): 53–59. doi: 10.3969/j.issn.1000-0755.2018.07.015
[12]	MUNASINGHE L and ICHISE R. Time score: A new feature for link prediction in social networks[J]. IEICE TRANSACTIONS on Information and Systems, 2012, E95-D(3): 821–828. doi: 10.1587/transinf.E95.D.821
[13]	ZHANG Zhongbao, WEN Jian, SUN Li, et al. Efficient incremental dynamic link prediction algorithms in social network[J]. Knowledge-Based Systems, 2017, 132: 226–235. doi: 10.1016/j.knosys.2017.06.035
[14]	邓志宏, 老松杨, 白亮. 基于预测误差修正的时序链路预测方法[J]. 电子与信息学报, 2014, 36(2): 325–331. doi: 10.3724/SP.J.1146.2013.00657 DENG Zhihong, LAO Songyang, and BAI Liang. A temporal link prediction method based on link prediction error correction[J]. Journal of Electronics &Information Technology, 2014, 36(2): 325–331. doi: 10.3724/SP.J.1146.2013.00657
[15]	BLISS C A, FRANK M R, DANFORTH C M, et al. An evolutionary algorithm approach to link prediction in dynamic social networks[J]. Journal of Computational Science, 2014, 5(5): 750–764. doi: 10.1016/j.jocs.2014.01.003
[16]	CASADIEGO J, NITZAN M, HALLERBERG S, et al. Model-free inference of direct network interactions from nonlinear collective dynamics[J]. Nature Communications, 2017, 8(1): 2192. doi: 10.1038/s41467-017-02288-4
[17]	WANG Wenxu, LAI Yingcheng, and GREBOGI C. Data based identification and prediction of nonlinear and complex dynamical systems[J]. Physics Reports, 2016, 644: 1–76. doi: 10.1016/j.physrep.2016.06.004
[18]	HAN Xiao, SHEN Zhesi, WANG Wenxu, et al. Robust reconstruction of complex networks from sparse data[J]. Physical Review Letters, 2015, 114(2): 028701. doi: 10.1103/PhysRevLett.114.028701
[19]	尹赢, 张建朋, 吉立新, 等. 基于霍克斯点过程的动态网络表示学习方法[J]. 电子学报, 2020, 48(11): 2154–2161. doi: 10.3969/j.issn.0372-2112.2020.11.009 YIN Ying, ZHANG Jianpeng, JI Lixin, et al. Dynamic network representation learning based on Hawkes point process[J]. Acta Electronica Sinica, 2020, 48(11): 2154–2161. doi: 10.3969/j.issn.0372-2112.2020.11.009
[20]	LÜ Linyuan and ZHOU Tao. Link prediction in complex networks: A survey[J]. Physica A:Statistical Mechanics and its Applications, 2011, 390(6): 1150–1170. doi: 10.1016/j.physa.2010.11.027
[21]	LIU Shuxin, JI Xinsheng, LIU Caixia, et al. Extended resource allocation index for link prediction of complex network[J]. Physica A:Statistical Mechanics and its Applications, 2017, 479: 174–183. doi: 10.1016/j.physa.2017.02.078
[22]	TANG Jian, QU Meng, WANG Mingzhe, et al. LINE: Large-scale information network embedding[C]. The 24th International Conference on World Wide Web, Florence, Italy, 2015: 1067–1077. doi: 10.1145/2736277.2741093.
[23]	LESKOVEC J, KLEINBERG J, and FALOUTSOS C. Graph evolution: Densification and shrinking diameters[J]. ACM Transactions on Knowledge Discovery from Data, 2007, 1(1): 2–es. doi: 10.1145/1217299.1217301
[24]	LESKOVEC J, LANG K J, DASGUPTA A, et al. Community structure in large networks: Natural cluster sizes and the absence of large well-defined clusters[J]. Internet Mathematics, 2009, 6(1): 29–123. doi: 10.1080/15427951.2009.10129177
[25]	VISWANATH B, MISLOVE A, CHA M, et al. On the evolution of user interaction in Facebook[J]. Proceedings of the ACM Workshop on Online Social Networks, 2009, 39(4): 37–42.
[26]	KUNEGIS J. Konect: The Koblenz network collection[C]. Companion: The 22nd International Conference on World Wide Web, Rio de Janeiro, Brazil, 2013: 1343–1350.
[27]	PANZARASA P, OPSAHL T, and CARLEY K M. Patterns and dynamics of users’ behavior and interaction: Network analysis of an online community[J]. Journal of the American Society for Information Science and Technology, 2009, 60(5): 911–932. doi: 10.1002/asi.21015
[28]	LESKOVEC J and KREVL A. SNAP datasets: Stanford large network dataset collection[DB/OL]. http://snap.stanford.edu/data/, 2018.
[29]	IBRAHIM N M A and CHEN Ling. Link prediction in dynamic social networks by integrating different types of information[J]. Applied Intelligence, 2015, 42(4): 738–750. doi: 10.1007/s10489-014-0631-0
[30]	李聪, 季新生, 李海涛, 等. 一种基于节点匹配度的动态网络链路预测方法[J]. 网络与信息安全学报, 2020, 5(4): 325–331. doi: 10.11959/j.issn.2096-109x.2019xxx LI Cong, JI Xinsheng, LI Haitao, et al. A link prediction method for dynamic networks based on matching degree of nodes[J]. Chinese Journal of Network and Information Security, 2020, 5(4): 325–331. doi: 10.11959/j.issn.2096-109x.2019xxx
[31]	SANKAR A, WU Yanhong, GOU Liang, et al. DySAT: Deep neural representation learning on dynamic graphs via self-attention networks[C]. The 13th International Conference on Web Search and Data Mining, Houston, USA, 2020: 519–527.
[32]	LI Jinsong, PENG Jianhua, LIU Shuxin, et al. TSAM: Temporal link prediction in directed networks based on self-attention mechanism[EB/OL]. arXiv: 2008.10021, 2020.