MEASURING ACCESSIBILITY: A BIG DATA PERSPECTIVE ON UBER SERVICE WAITING TIMES

ABSTRACT This study aims to relate information about the waiting times of ride-sourcing services, with specific reference to Uber, using socioeconomic variables from São Paulo, Brazil. The intention is to explore the possibility of using this measure as an accessibility proxy. A database was created with the mean waiting time data per district, which was aggregated to a set of socioeconomic and transport infrastructure variables. From this database, a multiple linear regression model was built. In addition, the stepwise method selected the most significant variables. Moran's I test confirmed the spatial distribution pattern of the measures, motivating the use of a spatial autoregressive model. The results indicate that physical variables, such as area and population density, are important to explain this relation. However, the mileage of district bus lines and the non-white resident rate were also significant. Besides, the spatial component indicates a possible relation to accessibility.

摘要 本研究旨在结合巴西圣保罗的社会经济变量，分析网约车召车服务的等待时长相关信息，并以优步（Uber）为具体研究案例，探索将该等待时长指标作为可达性代理变量的可行性。本研究构建了一则数据库，收录各行政区的平均等待时长数据，并整合了一系列社会经济与交通基础设施变量。基于该数据库，本研究构建了多元线性回归模型；此外，本研究通过逐步回归法筛选出最具显著性的变量。莫兰I（Moran's I）检验验证了该类指标的空间分布格局，促使本研究采用空间自回归模型。研究结果显示，面积、人口密度等实体变量对阐释该关联机制具有重要作用；不过，行政区内公交线路总里程与非白人居民占比同样具有统计显著性。此外，空间成分指标也体现出其与可达性可能存在关联。