Scholarlink

<p>Ride-hailing services have become an essential part of urban transportation. However, concerns are growing regarding their negative impact on traffic congestion, particularly due to deadheading, the movement of ride-hailing vehicles without passengers. This study examines how built environment characteristics influence the transfer rate of empty cars, a key metric for capturing deadheading behavior. Using data from the Didi Chuxing GAIA Open Dataset and Baidu Maps API, we employed the Gradient Boosting Decision Tree (GBDT) model. We then visualized nonlinear relationships through univariate and bivariate partial dependence plots. Our findings indicate that destination characteristics, such as the density of Points of Interest (POIs) and road networks, play a significant role in influencing the transfer rate of empty cars. Additionally, we observed a nonlinear relationship between the built environment and the transfer rate. Specifically, when road network density at the destination exceeds 6 km per km<sup>2</sup>, the transfer rate shifts from increasing to decreasing. Areas within a 15-km radius of the city center play a key role in ride-hailing activities due to moderate congestion and sustained demand. Finally, the OD-based interaction analysis reveals that the transfer rate is shaped by spatial interaction: origin-side constraints can push drivers away, while destination-side advantages pull them in. These insights underscore the need for integrated spatial planning that promotes balanced distribution of infrastructure and considers both ends of a ride-hailing trip to improve overall efficiency.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Impact of built environment on deadheading in ride-hailing services: a GBDT model analysis

  • 期刊论文

  • 发表日期：

    2025年7月3日

查看全文
  • Shuai Ling,
  • Xiaohan Su,
  • Shoufeng Ma,
  • Xin Zhang,
  • Jiong Gao,
  • Yifan Fu,
  • Xuan Feng

摘要

Ride-hailing services have become an essential part of urban transportation. However, concerns are growing regarding their negative impact on traffic congestion, particularly due to deadheading, the movement of ride-hailing vehicles without passengers. This study examines how built environment characteristics influence the transfer rate of empty cars, a key metric for capturing deadheading behavior. Using data from the Didi Chuxing GAIA Open Dataset and Baidu Maps API, we employed the Gradient Boosting Decision Tree (GBDT) model. We then visualized nonlinear relationships through univariate and bivariate partial dependence plots. Our findings indicate that destination characteristics, such as the density of Points of Interest (POIs) and road networks, play a significant role in influencing the transfer rate of empty cars. Additionally, we observed a nonlinear relationship between the built environment and the transfer rate. Specifically, when road network density at the destination exceeds 6 km per km2, the transfer rate shifts from increasing to decreasing. Areas within a 15-km radius of the city center play a key role in ride-hailing activities due to moderate congestion and sustained demand. Finally, the OD-based interaction analysis reveals that the transfer rate is shaped by spatial interaction: origin-side constraints can push drivers away, while destination-side advantages pull them in. These insights underscore the need for integrated spatial planning that promotes balanced distribution of infrastructure and considers both ends of a ride-hailing trip to improve overall efficiency.