Nonlinear effect of station-area built environment on taxi-metro combined travel
-
摘要: 预处理出租车轨迹数据并提取了出租车行程数据；识别了地铁运营时段内出租车-地铁组合出行，分类为接入出行、接出出行，并划定了地铁站点的潜在影响范围；采用极限梯度提升模型分析了站域建成环境对出租车-地铁组合出行的非线性影响，并引入SHAP模型解释了非线性特征的重要性、方向及阈值现象；运用必要条件分析方法探索了站域建成环境要素对出租车-地铁组合出行的必要性。研究结果表明：站域建成环境对不同类别的出租车-地铁组合出行具有不同的贡献度，其中土地利用混合度、公交站点密度、路网密度、居住用地密度对出租车-地铁组合出行具有显著的必要性；早高峰期间，影响接入出行的重要要素依次是公交站点密度、居住用地密度和路网密度，重要度分别为33.38%、30.10%、19.33%，影响接出出行的重要要素依次是公交站点密度、居住用地密度和路网密度，重要度分别为41.48%、15.61%、14.41%；晚高峰期间，影响接入出行的重要要素依次是居住用地密度、路网密度和公交站点密度，重要度分别为34.13%、23.84%、23.13%；影响接出出行的重要要素依次是居住用地密度、公交站点密度和土地利用混合度，重要度分别为40.88%、20.32%、14.72%；公交站点密度的SHAP值呈现3个阶段变化，以每平方公里15和50个站点为阈值点；路网密度在14 km·km-2处为重要分界点，低于该值对租车-地铁组合出行为负向贡献，高于该值为正向贡献；居住用地密度超过0.45 m2·km-2时对租车-地铁组合出行为正向贡献且趋于稳定。Abstract: The taxi trajectory data was preprocessed and taxi trip data was extracted. Taxi-metro combined travel was identified and categorized into subway extending access (SE-access) trips and subway extending egress (SE-egress) trips, and potential influence areas of subway stations were identified. The extreme gradient boosting (XGBoost) model was employed to analyze the nonlinear effects of station-area built environment on taxi-metro combined travel. The Shapley additive explanations (SHAP) model was introduced to explain the nonlinear characteristics: importance, orientation, and threshold phenomena. The necessity of station-area built environment factors for taxi-metro combined travel was analyzed using necessary condition analysis method. Analysis results indicate that the station-area built environment contributes differently to various types of taxi-metro combined travel, with land use mix, bus station density, road network density, and residential land density significantly necessary for taxi-metro combined travel. During the morning peak hours, key factors influencing SE-access trips are bus station density, residential land density, and road network density, with importance values of 33.38%, 30.10%, and 19.33%, respectively. For SE-egress trips, key factors are bus station density, residential land density, and road network density, with importance values of 41.48%, 15.61%, and 14.41%, respectively. During the evening peak hours, key factors influencing SE-access trips are residential land density, road network density, and bus station density, with importance values of 34.13%, 23.84%, and 23.13%, respectively. For SE-egress trips, key factors are residential land density, bus station density, and land use mix, with importance values of 40.88%, 20.32%, and 14.72%, respectively. The SHAP value of bus station density changes with a three-stage trend, with 15 and 50 stations per square kilometer as threshold points. For road network density, 14 km·km-2 is a critical demarcation value, meaning that the road network density below this threshold negatively contributes to taxi-metro combined travel, while the road network density above this threshold contributes positively. When residential land density exceeds 0.45 m2·km-2, it positively contributes to taxi-metro combined travel and tends to remain stable.
-
图 6 基于SHAP值的早高峰接入出行影响因素分析
Figure 6. Analysis of influencing factors for access trip during morning peak hours based on SHAP values
图 7 基于SHAP值的晚高峰接入出行影响因素分析
Figure 7. Analysis of influencing factors for access trip during evening peak hours based on SHAP values
图 8 基于SHAP值的早高峰接出出行影响因素分析
Figure 8. Analysis of influencing factors for egress trip during morning peak hours based on SHAP values
图 9 基于SHAP值的晚高峰接出出行影响因素分析
Figure 9. Analysis of influencing factors for egress trip during evening peak hours based on SHAP values
图 10 建成环境要素对早高峰接入出行的必要性
Figure 10. Necessities of built environment on access trips during morning peak hours
图 11 建成环境要素对晚高峰接入出行的必要性
Figure 11. Necessities of built environment on access trips during evening peak hours
图 12 建成环境对早高峰接出出行的必要性
Figure 12. Necessities of built environment on egress trips during morning peak hours
图 13 建成环境对晚高峰接出出行的必要性
Figure 13. Necessities of built environment on egress trips during evening peak hours
图 14 贡献率前5个特征的SHAP依赖关系(早高峰接入出行)
Figure 14. SHAP dependences of top 5 contributing features (access trips during morning peak hours)
图 15 贡献率前5个特征的SHAP依赖关系(晚高峰接入出行)
Figure 15. SHAP dependences of top 5 contributing features (access trips during evening peak hours)
图 16 贡献率前5个特征的SHAP依赖关系(早高峰接出出行)
Figure 16. SHAP dependences of top 5 contributing features (egress trips during morning peak hours)
图 17 贡献率前5个特征的SHAP依赖关系(晚高峰接出出行)
Figure 17. SHAP dependences of top 5 contributing features (egress trips during evening peak hours)
表 1 出租车轨迹数据格式
Table 1. Data format of taxi trajectory
字段 数据类型 实例 备注 车牌号 文本 甘AXXXX 出租车的唯一标识 时间 20210402083526 2021年4月2日08:35:26 载客状态 空车 空载状态 方向角 正北 正北方向 经度 数字 36.30 单位为° 纬度 103.78 单位为° 瞬时速度 35 单位为km·h-1 里程数 100 单位为km 
下载: 导出CSV
表 2 接入出行、接出出行换乘距离计算结果
Table 2. Calculation results of transfer distance for access and egress trips
类型 方式 时段 回归模型 R2 Sig 换乘距离/m 工作日 接入出行 早高峰 y=-0.005 4+2.777×10-4x1-1.392×10-8x12 0.997 0.000 3 847 晚高峰 y=-0.006 6+2.782×10-4x1-1.389×10-8x12 0.998 0.000 3 830 接出出行 早高峰 y=-0.001 8+2.751×10-4x2-1.365×10-8x22 0.997 0.000 3 796 晚高峰 y=-0.003+2.804×10-4x2-1.385×10-8x22 0.997 0.000 3 957 节假日 接入出行 早高峰 y=-0.047 2+2.849×10-4x1-1.418×10-8x12 0.998 0.000 3 910 晚高峰 y=-0.002 6+2.774×10-4x1-1.403×10-8x12 0.999 0.000 3 775 接出出行 早高峰 y=-0.066+3.088×10-4x2-1.566×10-8x22 0.998 0.000 3 794 晚高峰 y=-0.545×10-4x2-1.132×10-8x22 0.998 0.000 4 080
下载: 导出CSV
表 3 建成环境评价指标
Table 3. Evaluation indicators of built environment
特征因子 特征指标 分析尺度 数据来源 密度 不同类型土地利用的密度 1 km×1 km 政府规划 多样性 土地利用混合度 1 km×1 km 根据土地利用类型数据计算 设计 路网密度 1 km×1 km 开放街道地图 公共交通邻近性 公交站点/线网密度 1 km×1 km 百度地图
下载: 导出CSV
表 4 变量描述与统计
Table 4. Variable descriptions and statistics
变量类型 变量 描述 最小值 最大值 均值 标准差 因变量 Na1 每个网格内工作日早高峰时段接入出行量/人次 0 521 40.604 62.849 Ne1 每个网格内工作日晚高峰时段接入出行量/人次 0 364 31.022 47.775 Na2 每个网格内工作日早高峰时段接出出行量/人次 0 393 31.483 51.274 Ne2 每个网格内工作日晚高峰时段接出出行量/人次 0 404 39.135 54.606 自变量 E 每个网格内的土地利用混合度 0 0.444 0.183 0.108 PRT 每个网格内居住用地密度/(m2·km-2) 0 0.946 0.209 0.212 PC 每个网格内商业用地密度/(m2·km-2) 0 0.389 0.027 0.058 PI 每个网格内工业用地密度/(m2·km-2) 0 0.969 0.084 0.166 PRN 每个网格内休闲用地密度/(m2·km-2) 0 0.464 0.020 0.054 PO 每个网格内其他用地密度/(m2·km-2) 0 0.568 0.036 0.089 PRD 每个网格内路网密度/(km·km-2) 0 25.416 9.298 5.296 PBS 每个网格内公交站点密度/(个·km-2) 0 115.000 15.163 20.650 PBL 每个网格内公交线路密度/(km·km-2) 0 86.416 10.290 14.046
下载: 导出CSV
表 5 XGBoost模型拟合优度
Table 5. Goodness of fit of XGBoost model
变量 R2 训练集 验证集 接入出行 早高峰 0.79 0.60 晚高峰 0.75 0.64 接出出行 早高峰 0.74 0.59 晚高峰 0.78 0.69
下载: 导出CSV
表 6 站域建成环境要素的充分性和必要性(接入出行)
Table 6. Sufficiency and necessity of feeder-related built environment elements (access trips)
变量 早高峰接入出行 是否充分必要？(早高峰) 晚高峰接入出行 是否充分必要？(晚高峰) 充分性/% 必要性 充分性/% 必要性 E 9.74 0.164 否 13.47 0.120 否 PBS 33.38 0.398 是 23.13 0.346 是 PRD 19.33 0.589 是 23.84 0.576 是 PC 7.43 0.004 否 5.42 0.004 否 PI 0.00 0.000 否 0 0.000 否 PRT 30.10 0.295 是 34.13 0.363 是 PRN 0.00 0.005 否 0.00 0.004 否 PO 0.00 0.000 否 0.00 0.000 否
下载: 导出CSV
表 7 站域建成环境要素的其充分性和必要性(接出出行)
Table 7. Sufficiency and necessity of feeder-related built environment elements (egress trips)
变量 早高峰接出出行 是否充分必要？(早高峰) 晚高峰接出出行 是否充分必要？(晚高峰) 充分性/% 必要性 充分性/% 必要性 E 7.30 0.167 否 14.72 0.103 否 PBS 41.48 0.352 是 20.32 0.343 是 PRD 14.41 0.574 是 14.58 0.561 是 PC 5.94 0.657 否 3.36 0.004 否 PI 12.10 0.000 否 4.44 0.000 否 PRT 15.61 0.434 是 40.88 0.380 是 PRN 3.15 0.005 否 1.69 0.005 否 PO 0.00 0.000 否 0.00 0.000 否
下载: 导出CSV
-
[1] JIN Tan-hua, CHENG Long, ZHANG Xu-cai, et al. Nonlinear effects of the built environment on metro-integrated ridesourcing usage[J]. Transportation Research Part D: Transport and Environment, 2022, 110: 103426. doi: 10.1016/j.trd.2022.103426 [2] AZIMI G, RAHIMI A, LEE M, et al. Mode choice behavior for access and egress connection to transit services[J]. International Journal of Transportation Science and Technology, 2021, 10(2): 136-155. doi: 10.1016/j.ijtst.2020.11.004 [3] WELCH T F, GEHRKE S R, WIDITA A. Shared-use mobility competition: a trip-level analysis of taxi, bikeshare, and transit mode choice in Washington DC[J]. Transportmetrica A: Transport Science, 2020, 16(1): 43-55. doi: 10.1080/23249935.2018.1523250 [4] ZHAO Peng-jun, LI Sheng-xiao. Bicycle-metro integration in a growing city: the determinants of cycling as a transfer mode in metro station areas in Beijing[J]. Transportation Research Part A: Policy and Practice, 2017, 99: 46-60. doi: 10.1016/j.tra.2017.03.003 [5] FAN Yi-chuan, ZHENG Si-qi. Dockless bike sharing alleviates road congestion by complementing subway travel: evidence from Beijing[J]. Cities, 2020, 107: 102895. doi: 10.1016/j.cities.2020.102895 [6] WANG Jue-yu, CAO Xin-yu. Exploring built environment correlates of walking distance of transit egress in the Twin Cities[J]. Journal of Transport Geography, 2017, 64: 132-138. doi: 10.1016/j.jtrangeo.2017.08.013 [7] AZIZ A, NAWAZ M, NADEEM M S, et al. Examining suitability of the integrated public transport system: a case study of Lahore[J]. Transportation Research Part A: Policy and Practice, 2018, 117: 13-25. doi: 10.1016/j.tra.2018.08.003 [8] ZHANG Xiao-hu, XU Yang, TU Wei, et al. Do different datasets tell the same story about urban mobility—a comparative study of public transit and taxi usage[J]. Journal of Transport Geography, 2018, 70: 78-90. doi: 10.1016/j.jtrangeo.2018.05.002 [9] YANG Min, ZHAO Jing-yao, WANG Wei, et al. Metro commuters' satisfaction in multi-type access and egress transferring groups[J]. Transportation Research Part D: Transport and Environment, 2015, 34: 179-194. doi: 10.1016/j.trd.2014.11.004 [10] YANG L C, CHAU K W, SZETO W Y, et al. Accessibility to transit, by transit, and property prices: spatially varying relationships[J]. Transportation Research Part D: Transport and Environment, 2020, 85: 102387. doi: 10.1016/j.trd.2020.102387 [11] ZGHEIB N, ABOU-ZEID M, KAYSI I. Modeling demand for ridesourcing as feeder for high capacity mass transit systems with an application to the planned Beirut BRT[J]. Transportation Research Part A: Policy and Practice, 2020, 138: 70-91. doi: 10.1016/j.tra.2020.05.019 [12] 陈启香, 吕斌, 陈喜群, 等. 空间异质性建成环境对出租车与地铁竞合关系的影响[J]. 交通运输系统工程与信息, 2022, 22(3): 25-35.CHEN Qi-xiang, LYU Bin, CHEN Xi-qun, et al. Impacts of built environment on competition and cooperation relationship between taxi and subway considering spatial heterogeneity[J]. Journal of Transportation Systems Engineering and Information Technology, 2022, 22(3): 25-35. (in Chinese) [13] 周瑾. 基于轨道交通车站的巡游定制公交线路效率与配置研究[D]. 北京: 北京交通大学, 2023.ZHOU Jin. Research on line efficiency and configuration of cruise customized bus based on rail transit station[D]. Beijing: Beijing Jiaotong University, 2023. (in Chinese) [14] 丁川. 考虑空间异质性的城市建成环境对交通出行的影响研究[D]. 深圳: 哈尔滨工业大学(深圳), 2014.DING Chuan. Research on impacts of urban built environment on travel accounting for spatial heterogeneity[D]. Shenzhen: Harbin Institute of University, Shenzhen, 2014. (in Chinese) [15] 甘佐贤. 建成环境对城市轨道交通客流及出行特征的影响机理研究[D]. 南京: 东南大学, 2019.GAN Zuo-xian. Impact of built environment on the passenger flows and transfer behavior of urban rail transit[D]. Nanjing: Southeast University, 2019. (in Chinese) [16] LIU Qian, DING Chuan, CHEN Peng. A panel analysis of the effect of the urban environment on the spatiotemporal pattern of taxi demand[J]. Travel Behaviour and Society, 2020, 18: 29-36. doi: 10.1016/j.tbs.2019.09.003 [17] SABOURI S, PARK K, SMITH A, et al. Exploring the influence of built environment on Uber demand[J]. Transportation Research Part D: Transport and Environment, 2020, 81: 102296. doi: 10.1016/j.trd.2020.102296 [18] LIU Xing-hua, FAN Jing, LI Ye, et al. Analysis of integrated uses of dockless bike sharing and ridesourcing with metros: a case study of Shanghai, China[J]. Sustainable Cities and Society, 2022, 82: 103918. doi: 10.1016/j.scs.2022.103918 [19] GUO Y Y, HE S Y. Built environment effects on the integration of dockless bike-sharing and the metro[J]. Transportation Research Part D: Transport and Environment, 2020, 83: 102335. doi: 10.1016/j.trd.2020.102335 [20] GUO Y Y, HE S Y. The role of objective and perceived built environments in affecting dockless bike-sharing as a feeder mode choice of metro commuting[J]. Transportation Research Part A: Policy and Practice, 2021, 149: 377-396. doi: 10.1016/j.tra.2021.04.008 [21] MA Qing-yu, XIN Yan-an, YANG Hong, et al. Connecting metros with shared electric scooters: comparisons with shared bikes and taxis[J]. Transportation Research Part D: Transport and Environment, 2022, 109: 103376. doi: 10.1016/j.trd.2022.103376 [22] NI Ying, CHEN Jia-qi. Exploring the effects of the built environment on two transfer modes for metros: dockless bike sharing and taxis[J]. Sustainability, 2020, 12(5): 2034-2054. doi: 10.3390/su12052034 [23] 王建军, 王赛, 宋明洋, 等. 大数据背景下城市建成环境对出行行为影响研究综述[J]. 长安大学学报(自然科学版), 2022, 42(1): 61-78.WANG Jian-jun, WANG Sai, SONG Ming-yang, et al. Review of research on impact of urban built environment on travel behavior in big data context[J]. Journal of Chang'an University (Natural Science Edition), 2022, 42(1): 61-78. (in Chinese) [24] 刘畅, 郭亮, 杨朔, 等. 建成环境对公共交通与小汽车出行方式选择的非线性影响——基于出行链视角[J]. 热带地理, 2024, 44(5): 891-905.LIU Chang, GUO Liang, YANG Shuo, et al. Nonlinear effects of the built environment on public transit and private vehicle model choice: a trip chain perspective[J]. Tropical Geography, 2024, 44(5): 891-905. (in Chinese) [25] 陈奕璠, 张步镐, 党振, 等. 建成环境对共享单车使用特征的非线性影响研究[J]. 交通运输系统工程与信息, 2024, 24(2): 217-224.CHEN Yi-fan, ZHANG Bu-hao, DANG Zhen, et al. Exploring nonlinear effects of built environment on dockless bike sharing usage[J]. Journal of Transportation Systems Engineering and Information Technology, 2024, 24(2): 217-224. (in Chinese) [26] DING Chuan, CAO Xin-yu, LIU Cao. How does the station-area built environment influence metrorail ridership? Using gradient boosting decision trees to identify non-linear thresholds[J]. Journal of Transport Geography, 2019, 77: 70-78. doi: 10.1016/j.jtrangeo.2019.04.011 [27] TU Mei-ting, LI Wen-Xiang, ORFILA O, et al. Exploring nonlinear effects of the built environment on ridesplitting: evidence from Chengdu[J]. Transportation Research Part D: Transport and Environment, 2021, 93: 102776. doi: 10.1016/j.trd.2021.102776 [28] LYU Tao, WANG Pei-rong, GAO Ya-nan, et al. Research on the big data of traditional taxi and online car-hailing: a systematic review[J]. Journal of Traffic and Transportation Engineering (English Edition), 2021, 8(1): 1-34. doi: 10.1016/j.jtte.2021.01.001 [29] 陈启香. 基于建成环境的出租车与地铁竞合机理研究[D]. 兰州: 兰州交通大学, 2023.CHEN Qi-xiang. Research on the coopetition mechanism between taxi and subway based on the built environment[D]. Lanzhou: Lanzhou Jiaotong University, 2023. (in Chinese) [30] 佟欣. 轨道站域空间组织对步行服务效率的影响研究——以深圳市典型站域为例[D]. 深圳: 哈尔滨工业大学(深圳), 2020.TONG Xin. Research on impacts of spatial organization of railway station area on walking service efficiency: a case study of Shenzhen[D]. Shenzhen: Harbin Institute of University, Shenzhen, 2020. (in Chinese) [31] SUN Jian, PENG Zhong-ren, CHEN Wei-ya, et al. Development of web-based transit trip-planning system based on service-oriented architecture[J]. Transportation Research Record, 2011, 2217(1): 87-94. doi: 10.3141/2217-11 [32] 王淑伟, 孙立山, 荣建. 北京市轨道站点吸引范围研究[J]. 交通运输系统工程与信息, 2013, 13(3): 183-188.WANG Shu-wei, SUN Li-shan, RONG Jian. Catchment area analysis of Beijing transit stations[J]. Journal of Transportation Systems Engineering and Information Technology, 2013, 13(3): 183-188. (in Chinese) [33] WANG Zi-jia, CHEN Feng, XU Tian-kun. Interchange between metro and other modes: access distance and catchment area[J]. Journal of Urban Planning and Development, 2016, 142(4): 04016012. [34] LEE J, CHOI K, LEEM Y. Bicycle-based transit-oriented development as an alternative to overcome the criticisms of the conventional transit-oriented development[J]. International Journal of Sustainable Transportation, 2016, 10(10): 975-984. [35] 李甜. 考虑空间异质性的建成环境与交通系统运行状态关系建模及协调优化[D]. 镇江: 江苏大学, 2022.LI Tian. Research on the relationship model and coordination optimization between built environment and transportation state considering spatial heterogeneity[D]. Zhenjiang: Jiangsu University, 2022. (in Chinese) [36] GUO Yuan-yuan, YANG Lin-chuan, LU Yi, et al. Dockless bike-sharing as a feeder mode of metro commute? The role of the feeder-related built environment: analytical framework and empirical evidence[J]. Sustainable Cities and Society, 2021, 65: 102594. [37] CHEN Tian-qi, GUESTRIN C. XGBoost: a scalable tree boosting system[C]//ACM. 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2016: 785-794. [38] 谷岩岩. 大数据支持下的城市公共交通与土地利用互动研究[D]. 武汉: 武汉大学, 2020.GU Yan-yan. Research on interaction between urban public transportation and land use using big data[D]. Wuhan: Wuhan University, 2020. (in Chinese) [39] LUNDBERG S M, LEE S I. A unified approach to interpreting model predictions[J]. Advances in Neural Information Processing Systems, 2017, 2017: 4766-4775. [40] 吉淑娟. 建成环境对自行车使用和安全影响的异质性和非线性研究[D]. 西安: 长安大学, 2023.JI Shu-juan. Heterogeneous and non-linear effects of the built environment on bicycle use and bicycle safety[D]. Xi'an: Chang'an University, 2023. (in Chinese) [41] DUL J. Necessary condition analysis (NCA): logic and methodology of "necessary but not sufficient" causality[J]. Organizational Research Methods, 2016, 19(1): 10-52. -
点击查看大图
计量
- 文章访问数: 219
- HTML全文浏览量: 77
- PDF下载量: 23
- 被引次数: 0
