摘要
为捕捉由轨道交通站点周边建成环境与客流时变特征的互动关系而反映的站点类型差异,基于地铁刷卡数据与站点周边兴趣点(Point of Interest,POI)数据,分别通过客流时间序列分析和地理加权回归模型进行时空维度聚类变量提取.应用K-means++聚类算法将杭州地铁1、2、4号线站点划分为工作导向型、居住导向型、商业型以及工作-居住混合型4种类型.研究结果表明:该方法相对于传统K-means算法具有更优的性能表现,其中轮廓系数、Davies-Bouldin指数与Calinski-Harabaz指数等3项聚类评价指标的改善幅度分别为30.43%、10.51%、9.02%,因而能够准确识别时空视角下的轨道交通站点类型并反映其客流出行模式,进而为站点客流预测、站城一体化建设等后续研究提供分析依据.
To capture the differences in station types reflected by the relationship between the built environment around metro stations and time-varying characteristics of passenger flows,based on the metro swip card data and Point of Interest(POI)data around stations,the spatio-temporal clustering variables are extracted through the passenger flow time series analysis and geographically weighted regression model.By using K-means++clustering algorithm,stations in Hangzhou metro stations 1,2 and 4 are classified into four types:work-oriented stations,residential-oriented stations,commercial stations and mixed work-residence stations.The results show that compared with the traditional K-means algorithm,this method has better performance,with the improvement of 30.43%,10.51%and 9.02%in the three clustering evaluation indexes,including silhouette coefficient,Davies-Bouldin index and CalinskiHarabaz index.Therefore,this method can accurately identify the types of metro stations and corresponding passenger flow patterns from a spatio-temporal perspective,therefore providing reference for the subsequent study of station-level passenger flow forecasting and station-city integration development.
作者
李亮
赵星
张海燕
杜希旺
LI Liang;ZHAO Xing;ZHANG Haiyan;DU Xiwang(College of Civil Engineering and Transportation Engineering,Hohai University,Nanjing 210098,China)
出处
《北京交通大学学报》
CAS
CSCD
北大核心
2022年第4期31-42,共12页
JOURNAL OF BEIJING JIAOTONG UNIVERSITY
基金
江苏省自然科学基金(BK20211203)
中央高校基本科研业务费专项资金(2017B12714)。
