Abstract: Travel time is an essential factor for residents to choose their traffic modes. In order to explore the differences in the impact of urban built environment on bicycle travel time under different travel purposes, based on multi-source data such as Shantou residents travel survey data, road network data, and interest points, the impact of built environment on commuting and non-commuting bicycle travel time was analyzed by the extreme gradient boosting model and SHAP algorithm. The research indicates that the extreme gradient boosting model fits better than the multiple linear regression model. The importance of built environment variables on commuting and non-commuting bicycle travel time is different, and there is a nonlinear impact. The population density has the highest importance for commuting bicycle travel time, while the normalized difference vegetation index (NDVI) has the highest importance for non-commuting bicycle travel time. The influence of population density on the two types of bicycle travel time is V-shaped, while the density of bus-stops and road network is inverted V-shaped. NDVI, the number of office facilities and the mixing degree of land use have opposite effects on the two types of bicycle travel time.

Key words: traffic and transportation engineering; different trip purposes; urban built environment; bicycle travel time; nonlinear effects; extreme gradient boosting model

摘要： 出行时间是居民选择出行方式的重要因素，为探究不同出行目的下城市建成环境对自行车出行时间影响差异性，基于汕头市居民出行调查数据、路网数据、兴趣点等多源数据，运用极端梯度提升模型和SHAP算法，分析建成环境对通勤与非通勤自行车出行时间的影响。研究表明：极端梯度提升模型拟合效果优于多元线性回归模型；建成环境变量对通勤与非通勤自行车出行时间的重要度不同，且存在非线性影响；人口密度对通勤自行车出行时间的重要度最高，归一化植被系数（NDVI）对非通勤自行车出行时间的重要度最高；人口密度对两类自行车出行时间的影响呈“V”形，公交站点密度和路网密度则呈倒“V”形；NDVI、办公设施数量与土地利用混合度对两类自行车出行时间的影响相反。

关键词: 交通运输工程；不同出行目的；城市建成环境；自行车出行时间；非线性影响；极端梯度提升模型