联合子图模式和网络表征的路网链路预测模型  

Road Network Link Prediction Model Based on Subgraph Patterns and Network Embedding

作　　者：王斌[1] 李毅磊 盛津芳[1] 孙泽军 卢奔[1] WANG Bin;LI Yi-lei;SHENG Jin-fang;SUN Ze-jun;LU Ben(School of Computer Science and Engineering,Central South University,Changsha 410083,China;Department of Network Center,Pingdingshan University,Pingdingshan 467000,China)

机构地区：[1]中南大学计算机学院,长沙410083 [2]平顶山学院网络中心,河南平顶山467000

出　　处：《小型微型计算机系统》2019年第11期2357-2364,共8页Journal of Chinese Computer Systems

基　　金：国家重大科技专项项目(2017ZX06002005)资助

摘　　要：城市道路网络(简称路网)是一种结构复杂且高度稀疏的网络,对城市道路网络进行链路预测能够对城市结构变化进行合理预演,辅助城市设计者决策.本文针对路网特性提出了一种新的链路预测模型GRSC,该模型首先通过road2vec对路网进行网络表征,然后将子图模式和网络表征结果有机地结合起来,共同构建包含子图结构特征、游走距离特征的广义路网子图特征,最后训练logistic回归分类模型,用于路网链路预测.实验对比了GRSC模型和其它链路预测模型在不同国家、不同类型城市路网上的表现以及模型参数的变化对预测精度的影响,结果表明,GRSC在预测精度和稳定性方面都表现良好.Urban road network is a complex and highly sparse network. Link prediction based on urban road network can reasonably predict urban structural changes and assist urban designers in making decisions. General Road Subgraph Classification,a new link prediction model,is proposed for the characteristics of road network. The model firstly performs network embedding on the road network through road2 vec,and then organically combines the subgraph patterns and network embedding results to construct generalized road subgraph features,which including subgraph structure features and walking distance features. Finally,the logistic regression classification model is trained to act on the road network link prediction. The experiment compares the performance of GRSC model and other link prediction models in different types of urban road networks and compares the influence of model parameters on prediction accuracy. The results show that GRSC performs well in terms of prediction accuracy and stability.

关 键 词：链路预测 子图模式 网络表征 分类模型 城市路网 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]