一种基于线性规划的有向网络链路预测方法

李劲松; 彭建华; 刘树新; 季新生

doi:10.11999/JEIT190731

一种基于线性规划的有向网络链路预测方法

doi: 10.11999/JEIT190731

李劲松¹,
彭建华¹,
刘树新^1, ,,
季新生^{1, 2}

1.
中国人民解放军战略支援部队信息工程大学郑州 450001
2.
清华大学计算机科学与技术系北京 100084

基金项目: 国家自然科学基金(61803384)

详细信息

作者简介:
李劲松：男，1992年生，博士生，研究方向为复杂网络，链路预测，网络安全

彭建华：男，1966年生，研究员，研究方向为网络安全，云安全，复杂网络

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

季新生：男，1969年生，教授，研究方向为网络安全，云安全，复杂网络

通讯作者:
刘树新　liushuxin11@126.com

中图分类号: N92; TP393
计量
- 文章访问数: 1725
- HTML全文浏览量: 691
- PDF下载量: 109
- 被引次数: 0
出版历程
- 收稿日期: 2019-09-20
- 修回日期: 2020-05-25
- 网络出版日期: 2020-06-01
- 刊出日期: 2020-10-13

A Link Prediction Method in Directed Networks Via Linear Programming

1.
Information Engineering University, People’s Liberation Army Strategic Support Force, Zhengzhou 450001, China
2.
Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China

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

摘要

摘要: 大多数有向网络链路预测方法在计算节点相似性时没有充分考虑有向网络的结构特点，未区分不同有向邻居对连边形成具有的贡献差异，导致预测性能受到局限。鉴于此，该文提出一种基于线性规划的有向网络链路预测方法。该方法对3种有向邻居的信息贡献进行量化分析，结合结构特点建立线性规划模型，进而通过求解贡献矩阵的最优解构建相似性指标。9个真实有向网络中的实验结果表明，所提方法相比于9种现有方法在两种衡量标准下表现出较高的预测性能与良好的鲁棒性。
- 有向网络 /
- 链路预测 /
- 节点相似性 /
- 线性规划
Abstract: Most existing link prediction methods in directed networks fail to consider the structural properties of directed networks when calculating node similarity, nor do they differentiate the contributions of directed neighbors on link formation, resulting in the limitation on prediction performance. To solve these problems, a novel link prediction method in directed networks based on linear programming is proposed. The contributions of three types of directed neighbors are quantified, then the linear programming problem is established based on network topological property. The similarity index is deduced by solving the optimal solution of the linear programming problem. Experimental results on nine real-world directed networks show that the proposed method outperforms nine benchmarks on both accuracy and robustness under two evaluation metrics.
- Directed network /
- Link prediction /
- Node similarity /
- Linear programming

HTML全文

图 1 相似性矩阵的计算过程示意图

下载: 全尺寸图片幻灯片

图 2 相似性指标与对应邻接矩阵元素之间的关系示意图

下载: 全尺寸图片幻灯片

图 3 9组网络中ROC曲线对比图

下载: 全尺寸图片幻灯片

图 4 AUC随训练集划分比例变化曲线图

下载: 全尺寸图片幻灯片

图 5 AUC随参数$\alpha $变化曲线图

下载: 全尺寸图片幻灯片

表 1 网络数据基本统计参数

序号	名称	类型	$N$	$M$	$\left\langle k \right\rangle $	$\rho $(%)	$C$	$\left\langle d \right\rangle $	$\gamma $	$\kappa $
(1)	ADO	社交网络	2,539	12,969	10.22	38.8	14.2	5.30	0.251	8.25
(2)	HIG	社交网络	70	366	10.46	50.3	40.4	3.51	0.083	4.38
(3)	RES	社交网络	217	2,672	24.63	62.4	30.4	2.79	0.096	6.32
(4)	EMA	信息网络	1,005	25,571	50.89	71.1	25.7	3.01	–0.014	2.80
(5)	USA	交通网络	1,574	28,236	35.88	78.1	38.4	3.85	–0.113	1.85
(6)	OF	交通网络	2,939	30,501	20.76	97.2	25.5	5.19	0.051	1.74
(7)	CELE	生物网络	453	4,596	20.29	16.8	12.4	3.03	–0.226	2.62
(8)	FWFD	生态网络	128	2,137	33.39	2.9	31.4	1.88	–0.104	5.02
(9)	LAK	生态网络	183	2,494	27.26	4.09	33.2	2.65	–0.266	2.99

下载: 导出CSV

表 2 AUC结果对比

预测指标	ADO	HIG	RES	EMA	USA	OF	CELE	FWFD	LAK
DCN	0.716	0.860	0.889	0.949	0.969	0.968	0.804	0.745	0.942
DAA	0.714	0.862	0.895	0.953	0.972	0.969	0.807	0.749	0.940
DRA	0.716	0.861	0.895	0.956	0.973	0.971	0.811	0.752	0.939
DPA	0.680	0.621	0.650	0.887	0.953	0.924	0.810	0.854	0.946
Bifan	0.770	0.826	0.859	0.932	0.964	0.965	0.888	0.920	0.988
LP-0.001	0.781	0.882	0.899	0.954	0.976	0.984	0.860	0.769	0.964
Katz-0.001	0.877	0.883	0.898	0.954	0.976	0.984	0.874	0.778	0.970
MFI	0.872	0.845	0.792	0.772	0.908	0.975	0.836	0.734	0.947
LO-0.01	0.678	0.830	0.888	0.949	0.967	0.971	0.889	0.972	0.995
LO-0.001	0.679	0.830	0.868	0.956	0.977	0.974	0.892	0.934	0.990
LPD-0.01-1	0.803	0.898	0.917	0.956	0.984	0.991	0.899	0.956	0.994
LPD-0.001-1	0.812	0.883	0.902	0.969	0.968	0.990	0.886	0.907	0.988
LPD-max	0.814	0.898	0.919	0.969	0.984	0.993	0.907	0.973	0.996

下载: 导出CSV

参考文献(19)

REN Zhuoming, ZENG An, and ZHANG Yicheng. Structure-oriented prediction in complex networks[J]. Physics Reports, 2018, 750: 1–51. doi: 10.1016/J.PHYSREP.2018.05.002

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

王凯, 李星, 兰巨龙, 等. 一种基于资源传输路径拓扑有效性的链路预测方法[J]. 电子与信息学报, 2020, 42(3): 653–660. doi: 10.11999/JEIT190333

WANG Kai, LI Xing, LAN Julong, et al. A new link prediction method for complex networks based on topological effectiveness of resource transmission paths[J]. Journal of Electronics &Information Technology, 2020, 42(3): 653–660. doi: 10.11999/JEIT190333

LIU Shuxin, JI Xinsheng, LIU Caixia, et al. Similarity indices based on link weight assignment for link prediction of unweighted complex networks[J]. International Journal of Modern Physics B, 2017, 31(2): 1650254. doi: 10.1142/S0217979216502544

AGHABOZORGI F and KHAYYAMBASHI M R. A new similarity measure for link prediction based on local structures in social networks[J]. Physica A: Statistical Mechanics and its Applications, 2018, 501: 12–23. doi: 10.1016/J.PHYSA.2018.02.010

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

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

KATZ L. A new status index derived from sociometric analysis[J]. Psychometrika, 1953, 18(1): 39–43. doi: 10.1007/BF02289026

CHEBOTAREV P and SHAMIS E. The matrix-forest theorem and measuring relations in small social groups[J]. Automation and Remote Control, 1997, 58(9): 1505–1514.

ZHOU Tao, LÜ Linyuan, and ZHANG Yicheng. Predicting missing links via local information[J]. The European Physical Journal B, 2009, 71(4): 623–630. doi: 10.1140/EPJB/E2009-00335-8

LI Jinsong, PENG Jianhua, LIU Shuxin, et al. Link prediction in directed networks utilizing the role of reciprocal links[J]. IEEE Access, 2020, 8: 28668–28680. doi: 10.1109/ACCESS.2020.2972072

ZHANG Xue, ZHAO Chengli, WANG Xiaojie, et al. Identifying missing and spurious interactions in directed networks[J]. International Journal of Distributed Sensor Networks, 2015, 11(9): 507386. doi: 10.1155/2015/507386

WANG Xiaojie, ZHANG Xue, ZHAO Chengli, et al. Predicting link directions using local directed path[J]. Physica A: Statistical Mechanics and its Applications, 2015, 419: 260–267. doi: 10.1016/J.PHYSA.2014.10.007

BÜTÜN E and KAYA M. A pattern based supervised link prediction in directed complex networks[J]. Physica A: Statistical Mechanics and its Applications, 2019, 525: 1136–1145. doi: 10.1016/J.PHYSA.2019.04.015

SALHA G, LIMNIOS S, HENNEQUIN R, et al. Gravity-inspired graph autoencoders for directed link prediction[C]. The 28th ACM International Conference on Information and Knowledge Management, Beijing, China, 2019: 589–598. doi: 10.1145/3357384.3358023.

GUNDALA L A and SPEZZANO F. Estimating node indirect interaction duration to enhance link prediction[J]. Social Network Analysis and Mining, 2019, 9(1): 17. doi: 10.1007/s13278-019-0561-2

PECH R, HAO D, LEE Y L, et al. Link prediction via linear optimization[J]. Physica A: Statistical Mechanics and Its Applications, 2019, 528: e121319. doi: 10.1016/j.physa.2019.121319

ZHANG Qianming, LÜ Linyuan, WANG Wenqiang, et al. Potential theory for directed networks[J]. PLoS One, 2013, 8(2): e55437. doi: 10.1371/JOURNAL.PONE.0055437

KUNEGIS J. KONECT network dataset[EB/OL]. http://konect.uni-koblenz.de/networks/, 2017.

施引文献

资源附件(0)

访问统计