突发事件影响下的城市居民出行活动时空模式研究  

Spatiotemporal patterns of urban residents’travel activities under the impact of emergencies

作　　者：何惠雨 付晓 吕启航 HE Huiyu;FU Xiao;LYU Qihang(School of Transportation,Southeast University,Nanjing 211189,China;Key Laboratory of Transport Industry of Comprehensive Transportation Theory(Nanjing Modern Multimodal Transportation Laboratory),Ministry of Transport,Nanjing 210000,China)

机构地区：[1]东南大学交通学院,南京211189 [2]交通运输部综合交通运输理论交通运输行业重点实验室(南京现代综合交通实验室),南京210000

出　　处：《时空信息学报》2024年第2期259-268,共10页JOURNAL OF SPATIO-TEMPORAL INFORMATION

基　　金：国家自然科学基金项目(42261144745);教育部人文社会科学研究项目(21YJC790030)。

摘　　要：在人类社会的发展过程中,突发事件常常引发人们生活和行为的急剧变化,并可能对其产生持续性的影响。目前相关研究多为居民出行活动的总体趋势和整体特征,而在细分层面分析出行活动在时空维度上差异性的研究较少,且存在时空维度分离、缺乏整体性的问题。本文以美国旧金山新型冠状病毒感染流行为例,采用共享单车出行数据、兴趣点数据等进行长时间跨度的研究,利用k均值聚类和潜在狄利克雷分配模型,挖掘突发事件前后居民出行时空模式的特征及变化。结果表明:①事件暴发后,居民不同目的的出行活动时空模式发生了显著变化,通勤及娱乐出行的比例大幅下降,居民尝试用聚集程度较小的户外休闲娱乐代替聚集性娱乐,生活必需品购买及处理个人事务的出行比例上升,医疗需求大幅增加且该类型出行的早高峰开始时间提前。②随着时间的推移,突发事件的影响逐渐降低,人们的出行活动时空模式逐渐恢复至事件前的状态。研究成果可深化对风险和不确定性的认知,建立更全面的时空知识服务体系,为城市管理部门制定合理的应急管理策略提供参考。Throughout the course of human societal development,emergencies often trigger profound changes in individuals’lives and behaviors,leaving a lasting impact.While existing research primarily focuses on overarching trends and general characteristics of residents’travel activities,there remains a paucity studies delving into the finer spatiotemporal differences of such activities.Moreover,these studies often grapple with issues concerning disconnected spatiotemporal dimensions and lack comprehensiveness.This study employs the COVID-19 pandemic as a case study and conducts an extensive analysis spanning a significant period.Utilizing bike-sharing data and point of interest(POI)data,the study represents bike-sharing destination sites as feature vectors composing the proportions of various categories of POI points in their proximity.Three traditional clustering algorithms are compared,and the k-means algorithm,demonstrating superior performance,is utilized to cluster the sites.Subsequently,the study explores the land use characteristics of the bike-sharing sites obtained from each cluster.The clustering label of the destination site is used as the spatial feature of the trip,and when combined with the travel time feature,converts the trip record of the shared bicycle into the trip label containing spatial-temporal information.The entire dataset of trip label corpora is then used to train the latent Dirichlet allocation(LDA)model.This process yields the topic distribution for each document,i.e.,the topic distribution of the set of resident travel tags divided by month,and the word distribution for each topic,i.e.,the distribution of travel tags for each topic.The study observes the trend of topic distribution over time,with particular focus on the dominant topics prevalent during different periods.Based on the distribution of dominant topics’travel tags,the study explores the characteristics and changes in spatiotemporal patterns of resident travel before and after the emergency event.The results indicate significa

关 键 词：突发事件 出行活动 兴趣点 出行目的 聚类分析 时空模式 潜在狄利克雷分配模型 

分 类 号：U491[交通运输工程—交通运输规划与管理]