基于语义分析的网约车与地铁换乘特征研究

于泳波1，2，侯佳1，2

1．南京市城市与交通规划设计研究院股份有限公司，南京 210018

2．江苏省交通大数据与仿真平台技术工程研究中心，南京 210018

【摘要】为提升城市公交吸引力，在对网约车依赖较高区域和时段内增设公交服务以方便乘客与地铁接驳，需要了解乘客使用网约车与地铁接驳的特征。基于南京市网约车订单数据，通过语义分析技术，设定与地铁换乘接驳地点信息语义规则，设置与地铁站相关的5个基础词向量，再将每个地点抽象成词向量，通过计算与基础词向量的余弦相似度，提取与地铁换乘接驳的网约车出行订单信息。使用南京市居民出行调查数据和地铁AFC数据，以及手机信令数据校核识别结果，表明识别结果在可靠的范围内。研究表明，使用网约车与地铁接驳时，接驳距离为1～4km的行程数最多，约占全部行程的67%。早高峰期间，对于时间要求较高的通勤出行，居民更愿意通过打车的方法更快地到达地铁站。而晚高峰时段，部分从地铁站回家的居民依然会选择打车的方式，但打车的次数、比例均低于早高峰从家去地铁站的情形。研究成果可支撑特定区域和时段内，公交与地铁短途接驳线路规划与时刻表设计。

【关键词】城市交通；语义分析；网约车订单数据；换乘识别；客流来源去向分布

Transfer Characteristics of Online Car-hailing between Metro Based on Semantic Analysis

Yu Yongbo1,2,Hou Jia1,2

1．Nanjing Institute of City & Transport Planning Co.，Ltd，Nanjing 210018

2．Jiangsu Transportation Big Data and Simulation Platform Technology Engineering Research Center，Nanjing 210018

Abstract: In order to enhance the attractiveness of urban public transport, public transport services should be added in areas and time periods with high dependence on online car Hailing to facilitate the connection between passengers and subway. Based on Nanjing online car hailing order data, through semantic analysis technology, the semantic rules of subway transfer and connection location information were set, and five basic word vectors related to subway station were set. Then each location was abstracted into a word vector, and cosine similarity was calculated with the basic word vector to extract the transfer connection with subway online car hailing travel order information. Nanjing residents travel survey data and subway AFC data, as well as mobile phone signaling data was used to verify the identification results, it shows that the identification results are in a relia-ble range. It is concluded that when using the online car hailing to connect with the subway, the number of trips with the connection distance of 1-4km is the most, accounting for about 67% of the total journey. According to the analysis of the characteristics of different time periods and starting and ending points, it is concluded that residents are more willing to take a taxi to get to the subway station faster in the morning peak hours for commuters with higher time requirements. In the evening rush hour, some residents who come home from the subway station will still choose to take a taxi, but the number and proportion of taking a taxi are lower than that in the morning peak. The research results can support the planning and timetable design of bus and subway short distance connection in specific areas and time periods.

Key words：urban transportation;semantic analysis;online car-hailing data;transfer recognition;source and destination distribution of passenger flow