基于时空地理加权回归的共享单车需求影响因素分析  

Geographically and temporally weighted regression for modeling spatio-temporal variation in dockless bikeshare usage demand

作　　者：马新卫 季彦婕[1] 金雨川 金雪 MA Xin-wei;JI Yan-jie;JIN Yu-chuan;JIN Xue(School of Transportation,Southeast University,Nanjing 211189,China)

机构地区：[1]东南大学交通学院,南京211189

出　　处：《吉林大学学报（工学版）》2020年第4期1344-1354,共11页Journal of Jilin University:Engineering and Technology Edition

基　　金：国家重点研发计划项目(2018YFB1600900);国家自然科学基金项目(51561135003,51338003);东南大学优秀博士学位论文培育基金项目(YBJJ1842)。

摘　　要：基于南京市多源数据对共享单车出行需求进行了分析。构建了时空地理加权回归模型,揭示了公共自行车使用量、土地利用与社会经济人口因素在时间和空间上对共享单车出行需求的影响。研究表明:时空地理加权回归模型的拟合结果显著优于最小二乘回归模型与地理加权回归模型;公共自行车的使用会促进共享单车的需求,特别是在早晚高峰时段和郊区;娱乐型兴趣点在中心城区会促进共享单车的需求,在郊区则抑制共享单车的需求;老年人与小汽车用户不倾向使用共享单车,但在良好骑行环境且租赁自行车理念很好普及的区域更容易选择共享单车出行。To better understand the usage of dockless bikeshare,this paper investigates the influence of the docked bikeshare fleets,socio-demographic factors and land use factors on user demand of dockless bikeshare. To this end,the geographically and temporally weighted regression(GTWR) model was adopted as its ability to simultaneously incorporate spatial and temporal heterogeneity into dockless bikeshare data analysis,which outperforms the ordinary least squares(OLS)regression or geographically weighted regression(GWR)significantly in terms of model fit. Then,the spatial and temporal variations of the estimated coefficients were visualized and analyzed. Results show that docked bikeshare fleets promote dockless bikeshare usage,especially during rush hours and in the suburban areas. Entertainment POI(Point of Interest)promotes the dockless bikeshare usage in the urban areas during afternoon periods at weekends. The elderly groups and car owners prefer to use dockless bikeshare in the areas where bike infrastructure is well-developed. This study can offer meaningful implications for policy makers and dockless bike-sharing companies to improve dockless bike-sharing systems.

关 键 词：交通运输系统工程 共享单车 时空差异 土地利用 出行需求 时空地理加权回归模型 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]