During a pandemic or natural disaster, people may alter transit usage behavior due to perception of changes in the environment. To effectively respond to these crises, it is important for governments and public transit agencies to understand when these changes occurred and how they were affected by relevant policies and responsive strategies. In this study, we develop a methodological framework based on Bayesian online changepoint detection (BOCD) to identify the occurrence time, direction, and persistency of changes in individual-level transit usage. We demonstrate the effectiveness of this framework in informing government decision-making in the context of COVID-19. Using Jeju Island, South Korea as a case study, we apply the framework over a nearly two-year smart card dataset collected from the beginning of 2019 till nine months into the pandemic. By focusing on frequent transit users, we detect when these users significantly changed their transit usage frequency during the pandemic and identify several types of users who experienced different behavior change patterns. Besides demonstrating the great heterogeneity in individual-level behavior changes, we perform a regression analysis to further understand how these changes were affected by key government policies (e.g., Risk alert, Social distancing, Public transit policy, and Eased social distancing). Our results suggest that only certain sets of policies appear to have significant effects. In particular, introducing Risk alert would cause a 277% to 317% increase in the number of users who reduced transit usage frequency. Policies that eased social distancing, though, would cause a 134% to 155% increase in the number of users with travel frequency increase. The proposed BOCD framework enables a scalable solution to identifying and understanding changes of individual transit behavior. The methodology and findings are beneficial for developing targeted policies and interventions to facilitate daily travel and public transit operations during public health crises.

中文翻译：

---

评估流行病相关政策对个人公共交通出行模式的影响：基于贝叶斯在线变化点检测的框架

在大流行或自然灾害期间，人们可能会由于对环境变化的感知而改变交通使用行为。为了有效应对这些危机，政府和公共交通机构必须了解这些变化何时发生以及相关政策和应对策略如何影响这些变化。在本研究中，我们开发了一种基于贝叶斯在线变化点检测（BOCD）的方法框架，以识别个人层面交通使用变化的发生时间、方向和持续性。我们证明了该框架在为政府决策提供有关 COVID-19 的信息方面的有效性。我们以韩国济州岛为案例研究，将该框架应用于从 2019 年初到大流行九个月期间收集的近两年智能卡数据集。通过关注经常乘坐公共交通的用户，我们可以检测到这些用户在大流行期间何时显着改变了其公共交通使用频率，并确定了经历不同行为改变模式的几种类型的用户。除了证明个人层面行为变化的巨大异质性之外，我们还进行了回归分析，以进一步了解这些变化如何受到关键政府政策（例如风险警报、社交距离、公共交通政策和放松社交距离）的影响。我们的结果表明，只有某些政策集似乎具有显着效果。特别是，引入风险警报将导致减少公交使用频率的用户数量增加 277% 至 317%。不过，放松社交距离的政策将导致随着出行频率增加，用户数量增加 134% 至 155%。所提出的 BOCD 框架提供了一个可扩展的解决方案来识别和理解个人交通行为的变化。该方法和研究结果有利于制定有针对性的政策和干预措施，以促进公共卫生危机期间的日常出行和公共交通运营。