地理科学 ›› 2016, Vol. 36 ›› Issue (11): 1679-1687.doi: 10.13249/j.cnki.sgs.2016.11.010

• • 上一篇    下一篇

交通流网络的时空特征解析——基于张量分解方法视角

柯文前1,2,3(), 陆玉麒3, 朱宇1,2, 陈伟4, 杨青3   

  1. 1.福建师范大学地理研究所,福建 福州 350007
    2.湿润亚热带生态地理过程教育部重点实验室,福建 福州 350007
    3.南京师范大学地理科学学院,江苏 南京 210023
    4.中国科学院地理科学与资源研究所,北京 100101
  • 收稿日期:2015-12-27 修回日期:2016-02-08 出版日期:2016-11-10 发布日期:2016-11-10
  • 作者简介:

    作者简介:柯文前(1988-),男,福建晋江人,博士后,主要从事空间结构与交通地理学研究。E-mail:wqke2005@163.com

  • 基金资助:
    国家自然科学基金项目(41430635,41501178)、中国博士后科学基金项目(2016M590588)、福建省自然科学基金项目(2016J05093)资助

Traffic Flow Network Spatio-temporal Characteristics Analysis: Basis on the Tensor Decomposition Method Perspective

Wenqian Ke1,2,3(), Yuqi Lu3, Yu Zhu1,2, Wei Chen4, Qing Yang3   

  1. 1. Institute of Geography, Fujian Normal University, Fuzhou 350007, Fujian, China
    2. Key Laboratory of Humid Subtropical Eco-geographical Process (Fujian Normal University), Ministry of Education, Fuzhou 350007, Fujian, China
    3. College of Geography Science, Nanjing Normal University, Nanjing 210023, Jiangsu, China
    4. Institute of Geographic Sciences and Natural Resources Research, Beijing 100101, China
  • Received:2015-12-27 Revised:2016-02-08 Online:2016-11-10 Published:2016-11-10
  • Supported by:
    National Natural Science Foundation of China (41430635, 41501178), China Postdoctoral Science Foundation (2016M590588), Natural Science Foundation of Fujian Province (2016J05093).

摘要:

基于非负张量分解方法(NTF方法),依托 2004年1月至2012年12月江苏省高速公路网络的三维OD海量交通流数据,展开网络时空特征的解析、提取与挖掘,尝试解决现有数据分析方法无法有效解析交通流网络时空演化的动态性与多维性问题。结果表明:重构网络对原始网络的空间构型与格局具有很好的重现能力,基本刻画了原始网络的倒“不”字型空间结构;分解的网络可提取出基本不变型、渐变型和突变型3类时变规律,且每类时间特征有各自耦合对应的局部空间格局,体现内在组织的时空统一性;倒“不”字型空间结构由分解的沿沪宁线横向子网络叠加过润扬大桥沿扬溧线、过江阴大桥沿京沪线和过苏通大桥沿沈海线的三大纵向子网络等具有明确地理含义的多个局部空间系统共同构成,体现全局由局部组成的特性。

关键词: 张量分解方法, 时空特征, 交通流网络, 高速公路, 江苏省

Abstract:

The present researches on the spatio-temporal characteristics of the evolution of traffic flow network have paid less attention to the high-dimensional features and dynamic features. This article aims to analyze the spatio-temporal structure of expressway traffic network in Jiangsu base on the decomposition and reconstruction of the non-negative tensor factorization (NTF) method, and supported by a three-dimension matrix of “Origin-Destination-Time”, which consists of expressway traffic flow between 59 counties and time-evolving series from January 2004 to December 2012. The conclusions can be drawn as follows: 1) The reconstructed network has a good ability to reproduce the spatial construction and network pattern of original network. The spatial structure of both reconstructed network and original network are depicting a reverse “不” structure. 2) With the NTF method, there are three time-varying patterns at the study period which are essentially immutable pattern, gradually rising gradual pattern and rapidly rising mutational pattern. Each time-varying pattern has its own coupling and corresponding local spatial pattern respectively, which reflects the internal organization of spatio-temporal unity. 3) The reversed “不” spatial structure of the origin network consists of four clear geographical meaning systems. The Shanghai-Nanjing expressway connecting network acts as the horizontal axis of reversed “不” structure, the Jiangyin Changjiang River Highway Bridge along Beijing-Shanghai expressway connecting network acts as a vertical axis of reversed “不” structure, the Sutong Changjiang River Highway Bridge along Shenyang-Haikou expressway acts as the other vertical axis of reversed “不” structure, and the Runyang Changjiang River Highway Bridge acts as the last vertical axis of reversed “不” structure. The analysis reveals that each system is a local mode of the whole spatio-temporal network.

Key words: tensor decomposition method, spatio-temporal characteristics, traffic flow network, expressway, Jiangsu province

中图分类号: 

  • K901