热带地理 ›› 2017, Vol. 37 ›› Issue (1): 102-111.doi: 10.13284/j.cnki.rddl.002918

• 论文 • 上一篇    下一篇

基于地铁客流的广州地铁站点类型识别

谭章智1,李少英2,黎 夏1,刘小平1,陈逸敏1   

  1. (1.中山大学 地理科学与规划学院,广州 510275;2.广州大学 地理科学学院,广州 510006)
  • 出版日期:2017-01-05 发布日期:2017-01-05
  • 通讯作者: 李少英(1987―),女,广东汕头人,博士,主要研究方向为遥感与GIS应用、交通与土地利用互动模拟,(E-mail)lsy_0130@163.com。
  • 作者简介:谭章智(1988―),男,广东韶关人,博士研究生,主要研究方向为交通与土地利用互动关系,(E-mail)gistop@126.com
  • 基金资助:

    国家自然科学基金项目(41401432);广东省教育厅青年创新人才项目(2014KQNCX107);广东省自然科学基金博士启动项目(2015A030310288)

Clustering of Metro Stations in Guangzhou Based on Passenger Flow

TAN Zhangzhi1,LI Shaoying2,LI Xia1,LIU Xiaoping1,CHEN Yimin1   

  1. (1.School of Geography and Planning,Sun Yat-Sen University,Guangzhou 510275,China;2.School of Geographical Science,Guangzhou University,Guangzhou 510006,China)
  • Online:2017-01-05 Published:2017-01-05

摘要:

使用地铁刷卡数据对地铁客流进行研究时,面临数据量大、数据维度高、信息冗余度高等问题。传统方法主要通过专家经验建立指标进行分析,不能有效地利用原始数据,也容易受主观因素的影响。文章使用主成分分析方法(PCA)对广州地铁站点客流数据进行特征提取,用以提高特征选取过程的科学性,避免主观因素的影响,降低数据维度、消除冗余信息的过程中能够尽量保留原始数据的信息。研究表明:PCA提取出6个主成分,保留了原始数据91.41%的信息,前2个主成分代表了工作日地铁客流最重要的居住型站点客流特征和就业型站点客流特征。在此基础上利用K-均值聚类算法对广州地铁站点进行类型识别,识别出居住导向型、就业导向型、职住错位型等7类站点。通过分析不同站点的空间分布,发现广州地铁站点呈圈层式分布特征,站点类型随着到城市中心距离的增加而减少,反映了广州市城市功能的空间分布,城市中心区域站点类型多样,说明这一区域城市发展成熟、功能齐全;城市中心区域周围主要分布着职住错位型站点,城市发展较为成熟;而城市外围主要分布着居住导向型站点,承担城市的居住职能。

关键词: 地铁站点, 地铁客流, K-均值聚类算法, 主成分分析, 广州

Abstract:

Big data such as transportation card data, cell phone data provide data support for studies on trip characteristics analysis and city spatial structure. These studies require efficient methods to deal with the large volume, high dimension and information redundancy problems with big data. With the increase of data dimension, the complexity of data increases dramatically, making it impossible for human to understand the data and extract features based on expert knowledge. In this paper, principal component analysis (PCA) was used for dimensionality reduction of and feature extraction from passenger flow data of Guangzhou metro stations. The PCA process calculates the scoring coefficients for each component automatically without prior knowledge and eliminates the disturbance of subjective factors. Six principal components, extracted out of 36 variables, in this research kept 91.41% information of the original data. The first two components represented the passenger flow characteristics of residence-oriented stations and employment-oriented stations, respectively. K-means clustering of metro stations was performed based on the features extracted and 7 types of metro stations were recognized: residence-oriented, employment-oriented, spatially mismatched, mixed but more of residence-oriented, mixed but more of employment-oriented, comprehensive and traffic hub and entertainment stations. Among the 137 stations of Guangzhou Metro, over 85% were commuting-related, 46 stations were residence-oriented, 14 were employment-oriented, and 59 were of mixed nature of the two, with more of one or another. Further research on the spatial distribution of different clusters of the metro stations revealed a ring structure of Guangzhou City. The diversity and types of metro stations varied as the distance to city center increased, reflecting the spatial distribution of urban functions. Distribution of multi-types of metro stations such as employment-oriented, comprehensive and mixed but more of employment-oriented ones in the central city area such as Yuexiu District and Liwan District implied the high degree of development and the diversity of urban functions of these areas. Employment-oriented and comprehensive stations were intensely distributed in Tianhe District, making it the core administrative area and business area and the new city center of Guangzhou City. The main types of metro stations in Haizhu District were spatially mismatched stations and mixed but more of residence-oriented stations. In suburban areas such as Baiyun District and Panyu District, the main type of metro station was spatially mismatched stations, suggesting the decrease in the diversity of urban function. And only residence-oriented stations were found in the periphery areas of the city. Far away from the center of Guangzhou City, Nansha District and Foshan City developed into regional centers with some spatially mismatched stations distributed. The results of this study demonstrate that metro passenger flow data can not only reflect the spatial and temporal patterns of residents’ travel behavior, but also provide new data and a new perspective for urban spatial structure research.

Key words: metro stations, metro passenger flow, clustering, principal component analysis, Guangzhou Metro