文章编号： 1672-5328（2020）05-0054-07

阚长城1，马琦伟2，党安荣2，万涛3，高唱4
(1. 百度时代网络技术(北京)有限公司，北京100085；2. 清华大学建筑学院，北京100084；3. 天津市城市规划设 计研究院，天津300201；4.中国城市规划设计研究院，北京100037)

摘要： 通勤数据的传统获取手段存在成本高、覆盖面小、更新慢等问题，难以满足实时、高效监测 和管理的需求。基于百度地图时空大数据，综合应用多种机器学习方法，构建一套识别城市街区尺 度下通勤出行方式的技术框架，具有准确率高、覆盖面广、空间分辨率高等优势。挖掘北京市六环 高速公路以内各街区的通勤出行方式构成特征，结果显示各类交通方式的通勤出行比例相对均衡。 进而考察通勤出行方式的具体空间布局，探索其与路网密度、用地功能混合密度、公共交通设施服 务水平3 项建成环境因子之间的内在联系。对比分析结果验证了路网密度、用地功能混合密度与小 汽车通勤出行呈负相关关系，轨道交通服务水平对通勤出行的影响具有空间异质性。

关键词： 通勤交通方式；时空大数据；机器学习；绿色交通

中图分类号： U491

文献标识码：A

Commuting Travel Mode and Its Influencing Factors at Street Bock Scale: A Case Study in Beijing

Kan Changcheng1, Ma Qiwei2, Dang Anrong2,Wan Tao3, Gao Chang4
(1.Baidu.com Times Technology (Beijing) Co., Ltd., Beijing 100085, China; 2.School of Architecture, Tsinghua University, Beijing 100084, China; 3.Tianjin Urban Planning & Design Institute, Tianjin 300201, China; 4.China Academy of Urban Planning & Design, Beijing 100037, China)

Abstract： The disadvantages of the traditional commuting data collection methods are high cost, insufficient special coverage, and slow update, which makes it hard for meeting the real-time effective traffic monitoring and managing needs. Based on the Baidu Map's big spatial- temporal data, this paper develops a technical framework that uses the multiple machine learning methods to identify commuting travel mode at street block scale, which has high accuracy, wide coverage, and high spatial resolution. By analyzing the characteristics of commuting travel mode in different street blocks within the Sixth Ring freeway in Beijing, the results show that all types of commuting travel modes are relatively balanced. Based on the spatial distribution of commuting travel modes, the paper discusses the relationship among three built environment factors, such as roadway network density, mix land use density, and level of service of public transit facilities. The comparative analysis results show that roadway network density and mix land use density are negatively correlated with car commuting travel, and the impact of rail transit level of service on commuting travel is spatially heterogeneous.

Keywords： commuting travel modes; big spatial-temporal data; machine learning; green transportation