北京航空航天大学学报 ›› 2021, Vol. 47 ›› Issue (9): 1729-1738.doi: 10.13700/j.bh.1001-5965.2020.0335

• 论文 • 上一篇    下一篇

基于链路预测的未来新增航线发现

冯霞1,2, 王尧1,2   

  1. 1. 中国民航信息技术科研基地, 天津 300300;
    2. 中国民航大学 计算机科学与技术学院, 天津 300300
  • 收稿日期:2020-07-13 发布日期:2021-10-09
  • 通讯作者: 冯霞 E-mail:caucfx@126.com
  • 基金资助:
    国家自然科学基金(61502499);中央高校基本科研业务费专项资金(3122018C024);天津市自然科学基金(18JCYBJC85100)

Future air routes discovery based on link prediction

FENG Xia1,2, WANG Yao1,2   

  1. 1. Information Technology Research Base of Civil Aviation Administration of China, Tianjin 300300, China;
    2. School of Computer Science and Technology, Civil Aviation University of China, Tianjin 300300, China
  • Received:2020-07-13 Published:2021-10-09
  • Supported by:
    National Natural Science Foundation of China (61502499); the Fundamental Research Funds for the Central Universities (3122018C024); Tianjin Municipal Natural Science Foundation (18JCYBJC85100)

摘要: 针对新增航线发现研究中存在的航线选择主观化、网络信息挖掘不充分等问题,考虑航空运输网络的拓扑结构特征和节点(通航城市)层次属性,提出了一种基于链路预测的未来新增航线发现(NARP)模型。NARP模型提取局部封闭子图构建子图邻接矩阵,基于距离标记子图节点结构重要性,采用因子分析和层次聚类提取节点层次属性。在此基础上,融合子图结构和节点属性2类特征,采用深度图卷积神经网络(DGCNN)进行链路预测,实现新增航线发现。在中国航空运输网络实际运行数据上的实验结果表明:较之基准方法,NARP模型的预测准确率最高提升9.28%;在网络极度不完整时,预测准确率可以保持在80%左右;预测结果符合航空运输网络的实际演变情况。

关键词: 航空运输网络, 链路预测, 未来新增航线发现(NARP), 节点层次属性, 深度图卷积神经网络(DGCNN)

Abstract: In view of the problems of subjective route selection and insufficient network information mining in the research of new air routes discovery, and considering the topological structure characteristics and nodes (navigable cities) hierarchical attributes of air transport network, a New Air Routes Prediction (NARP) model based on link prediction is proposed. The NARP model extracted local enclosing subgraphs to construct subgraph adjacency matrices, marked the structural importance of nodes based on distance, and obtained nodes hierarchical attributes through factor analysis and hierarchical clustering. Then the two types of features of subgraph structure and nodes attributes were fused, and the Deep Graph Convolutional Neural Network (DGCNN) was used to perform link prediction to discover the future new air routes. The experimental results on the actual operation data of Chinese air transport network show that, compared with the benchmark algorithm, the prediction accuracy rate of NARP model is improved by 9.28% at most. When the network is extremely incomplete, the prediction accuracy rate can remain around 80%. The predicted results are in line with the actual evolution of air transport network.

Key words: air transport network, link prediction, New Air Routes Prediction (NARP), nodes hierarchical attributes, Deep Graph Convolutional Neural Network (DGCNN)

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发