申请同济大学工学博士学位论文


基于移动通信数据的城市居民活动特征及分类方法研究

 

 

 

所在院系:交通运输工程学院

学科专业:交通运输规划与管理

姓  名: 程小云

指导教师:杨东援 教授

二○一四年十二月

 

A dissertation submitted to

Tongji University in conformity with the requirements for

the degree of Doctor of Philosophy

Research on Travel Characteristics and

Classification of Urban Residents Based on Mobile Phone Data

 

 

School: School of Transportation Engineering

Major: Transport Planning and Management

Candidate: Xiaoyun Cheng

Supervisor: Prof. Yang Dong-yuan

December, 2014

 

摘  要

  快速城镇化、机动化及信息化加速了城市空间重组,改变了城市的空间格局,影响着在城市中活动的人的时空间行为。城市中各要素的空间布局,既是城市居民活动的主要载体,也是影响城市居民时空间行为的重要因素。因此,对城市环境与居民时空间行为相互影响关系的研究对城市规划和交通发展具有重要意义。
  互联网技术和信息通信技术(Information Communication Technology, ICT)的发展,对城市中居民的时空间行为数据提供了一种更有效的检测手段,更为城市空间组织与居民时空间行为影响关系的研究方法的变革提供了新契机。而如何清晰表达居民时空间行为、有效度量时空间行为特征及其差异性是城市环境与居民时空间行为相互关系研究的重要理论基础。因此,在信息数据日益多样化的趋势下,尝试运用描述性统计方法表征行为多维属性特征,采用关联分析等数据挖掘方法讨论时空间行为特征内在规律是城市时空间行为研究的重要方向。
  论文旨在通过对移动通信数据的挖掘分析,提出能够反映个体活动多维属性特征的城市居民时空间行为的表征方法,深入探讨不同城市空间要素布局下居民时空间行为的差异性,揭示城市交通的潜在问题。进而在宏观层面上,通过融合居民出行调查数据、整合专家意见,对所发现的交通问题进行解答。本文以上海市轨道交通7号线沿线上位于不同城市圈层的三个社区为研究案例,重点分析不同城市空间要素(重要锚点)的居民群体时空间行为差异性,揭示了在现有交通环境的实际需求下居民的集聚特征。
  文章首先针对这种新的数据源主要从数据特征、数据数量与质量的角度进行基本的数据描述与统计分析,表明该数据源在样本量、覆盖率及时间跨度上都对居民行为空间具有较强的连续说明能力。
  通过挖掘分析发现基于居民活动的时间序列位置信息数据,可从活动强度、复杂度、范围及与轨道交通的关系等角度提取个体的多维活动特征,构建描述个体时空间行为的多维属性分析框架。通过对比分析三个社区居民的时空间行为特征,表明该表征框架可清晰描述个体时空行为特征,有效度量具有不同主要锚点的群体时空行为特征差异性。
  通过观测具有相似时空行为特征的人群集聚情况,采用关联分析法从三个社区全体居民中识别出具有显著时空行为特征差异的三类主要人群。研究表明所识别的三类人群可有效表征不同社区居民的组成结构及特征。将这种识别方法作为一种新的不同于传统的以社会经济属性为依据的人群划分方法,能更客观的反映现有城市环境中居民个体对交通条件的实际需求。
  结合居民出行调查数据的分析结果,基于多源数据的决策层信息融合框架,通过综合专家组的经验评估,证实了居住地所在圈层是居民时空间行为差异性的重要影响因素,以及原动迁地对外围大型社区居民时空行为特征具有重要影响作用的决策分析,对研究城市交通的内在规律或机制提供了支持。
  通过研究发现,对位于城市不同圈层的三个社区,居民的时空间行为特征在不同社区间及相同社区内都具有显著差异,在讨论群体时空间行为特征时需采用不同的方法进行集计。以重要锚点为空间要素的集计分析,揭示了不同城市环境中居民时空间行为的差异性,并通过多源数据在决策层上的信息融合证实了居住地所在圈层是居民活动模式的重要影响因素。基于分析所得三类主要人群可更直观、更有效地表示一定城市空间内居民的组成结构及其特征,为需求分析提供一种更加灵活的集计方法。
  本研究成果对于城市时空间行为研究具有较好的可拓展应用性,通过与反映空间要素布局的时空数据相结合,能全面了解城市空间各要素布局与居民时空间行为的相互关系,有效管理、高效协调用户需求与交通供给之间的关系,提升规划者和决策者准确把握交通发展趋势与定位的能力,对于未来宏观决策分析具有广阔的应用前景。

关键词: 移动通信数据,时空间行为,空间活动模式,关联分析法,决策层信息融合

 

ABSTRACT

The rapid development of urbanization, motorization and informatization has been accelerating the reorganization of urban space, changing the urban spatial pattern and influencing the temporal and spatial behavior of residents lived in the city. The spatial distribution of urban elements is the major carrier of urban residents’ activities and the important factor of affecting the temporal and spatial behavior as well. Therefore, research on the interrelationship of urban environment and the residents’ temporal and spatial behavior plays a significant role in urban planning and transportation development.

The development of Information Communication Technology (ICT) provided an effective approach for the data collection of residents’ temporal and spatial behavior and an opportunity for the transformation of research methods on the interrelationship of urban environment and residents’ behavior. Then how to have a clear understanding of residents’ temporal and spatial behavior and how to effectively measure and differentiate the behavior features are the premise of the relevant researchs. Therefore, with the diversified data sources, trying to use descriptive statistical analysis to characterize the multidimensional attributes of residents’ behavior and adopt data mining method such as correlation analysis to discover the inherent law is a unique perspective for understanding the complex relationships between residents’ behavior and urban environments in space and time.

This paper aims at proposing a method of characterizating urban residents’ temporal and spatial behavior to reflect the multidimensional attributes of individual activity through mining mobile phone data. Then the differences of residents’ attributes under different urban spatial elements distribution are highlighted to find some potential traffic problems. Finally, these macroscopic traffic problems will be explained through fusing the travel behavior and attitude survey data and integrating expertise. Three communities along the rail transit line 7, located indifferent city circle-layers' region respectively in Shanghai are chosen as research areas. The difference of the collective temporal and spatial behavior with different urban spatial elements (important anchor point) is analyzed and then the characteristics of urban residents cluster are revealed in the contest of actual demand in existing traffic environment.

From data characteristics, data quantity and quality perspectives, this paper makes basic data description and statistical analysis. The results demonstrate the strong ability that the cellular network data can give a continuous representation for the temporal and spatial behavior in aspects of sample size, coverage and time span.

Through data mining and analysis, the multidimensional attributes of individual activity can be characterized in terms of activity intensity, complexity, scope and the relationship with rail transit. These characteristics extracted from the time series location information data can be used to build an analytical framework for describing individual temporal and spatial behavior. After comparison analysis, the results indicate that this framework can clearly descript the multidimensional attributes and effectively gauge the difference. By examining the cluster of residents with similar behavior profiles, three major groups are recognized among the population of these three communities via correlation analysis method. As a new classification different from the conventional method according to the social and economic attribute of people, this method can reflect the actual demand to traffic condition in the existing urban environment objectively.

Combining with the research findings of travel behavior survey data, it is proofed by the decision-level information fusion model that the city circle-layer of residence is an important factor influencing the residents’ temporal and spatial behavior and the old residential district plays vital role in the behavior of residents lived in large-scale residential community in suburban. These conclusions can provide support for exporing the inherent law and mechanism of urban transportation.

It can be concluded that people lived in different city circle-layer have significant discrepancy of temporal and spatial behavior both intra-and inter-community, which means studying on the collective behavior needs different aggregate methods for different purposes. Aggregating according to the main anchor point uncovered the temporal and spatial behavior discrepancy in different urban environment. And the importance of residence’s city circle-layer has been proofed by decision-level information fusion based on multi-source data. The recognized three main categories can illustrate the structural compositions of residents and their behavior features in a certain urban space more intuitively and effectively, which can be used as a more flexible aggregate method for demand analysis.

The research achievements in this paper have a great scalable application in urban temporal and spatial behavior research. By combining with spatial elements data, the interrelationship of urban environment and residents’ temporal and spatial behavior can be grasped to effectively manage and coordinate the relationship between traffic demand and supply and improve the capacity of planer and decision-maker to master the trend and direction of transportation development in future macro-decisive analysis.

Key Words:  Mobile phone data, Temporal and spatial behavior, Activity pattern, Correlation analysis method, Decision-level information fusion