Manley Ed, Zhong Chen, Batty Michael
Centre for Advanced Spatial Analysis, University College London, Gower Street, London, WC1E 6BT UK.
Transportation (Amst). 2018;45(3):703-732. doi: 10.1007/s11116-016-9747-x. Epub 2016 Nov 10.
New smart card datasets are providing new opportunities to explore travel behaviour in much greater depth than anything accomplished hitherto. Part of this quest involves measuring the great array of regular patterns within such data and explaining these relative to less regular patterns which have often been treated in the past as noise. Here we use a simple method called DBSCAN to identify clusters of travel events associated with particular individuals whose behaviour over space and time is captured by smart card data. Our dataset is a sequence of three months of data recording when and where individual travellers start and end rail and bus travel in Greater London. This dataset contains some 640 million transactions during the period of analysis we have chosen and it enables us to begin a search for regularities at the most basic level. We first define measures of regularity in terms of the proportions of events associated with temporal, modal (rail and bus), and service regularity clusters, revealing that the frequency distributions of these clusters follow skewed distributions with different means and variances. The analysis then continues to examine how regularity relative to irregular travel across space, demonstrating high regularities in the origins of trips in the suburbs contrasted with high regularities in the destinations in central London. This analysis sets the agenda for future research into how we capture and measure the differences between regular and irregular travel which we discuss by way of conclusion.
新的智能卡数据集为更深入地探索出行行为提供了新机会，其深度远超迄今所取得的任何成果。这项探索工作的一部分涉及测量此类数据中大量的常规模式，并相对于过去常被视为噪声的不太常规的模式来解释这些常规模式。在此，我们使用一种名为DBSCAN的简单方法来识别与特定个体相关的出行事件集群，智能卡数据记录了这些个体在空间和时间上的行为。我们的数据集是三个月的数据序列，记录了大伦敦地区个体旅行者铁路和公交出行的起止时间和地点。在我们选择的分析期间，这个数据集包含约6.4亿笔交易，这使我们能够在最基本的层面上开始寻找规律。我们首先根据与时间、模式（铁路和公交）以及服务规律集群相关的事件比例来定义规律度量，结果表明这些集群的频率分布遵循具有不同均值和方差的偏态分布。然后分析继续考察相对于跨空间的不规则出行的规律情况，结果显示郊区出行起点的规律性较高，而伦敦市中心目的地的规律性也较高。这项分析为未来研究设定了议程，即我们如何捕捉和衡量常规出行与不规则出行之间的差异，我们将在结论部分进行讨论。
