基于梯度提升决策树的重庆市轨道交通网络特征对站点客流的影响分析

Analysis of Effect of Chongqing Rail Transit Network Characteristics on Passenger Flow at Stations Based on Gradient Boosting Decision Trees

为了探讨轨道交通网络特征对站点客流的影响,基于重庆市轨道交通站点客流数据、POI数据、道路网矢量数据等多源数据,选择表征网络结构特征的中心性、连通性、集聚性等22个影响因子,采用普通最小二乘法和梯度提升决策树模型对客流进行拟合。结果表明,梯度提升决策树模型较普通最小二乘法的拟合度更好,其中站点效率是影响轨道站点客流最重要的因素,其次是临近中心性与站点重要性排序,对站点客流贡献度大于5%的因子共8个;同时对临近中心性、站点效率、公交站点密度等在内的9个影响因子进行独立效应分析,得到这些因子对站点客流均表现出非线性关系,且阈值效应显著。研究结果可为重庆市轨道交通资源的配置和轨道站点设置等提供一定的技术支持,也为其他城市轨道交通运营提供理论参考。

To explore the effect of rail transit network characteristics on passenger flow at stations,based on Chongqing’s multi-source data including passenger flow at rail transit stations,Point of Interest(POI)data,and road network vector data,this paper selected 22 influencing factors including centrality,connectivity,and clustering to represent network structure characteristics.Then,the ordinary least squares(OLS)and gradient boosting decision tree(GBDT)model were used in fitting the passenger flow.The results show that the fitting degree of the GBDT model is better than that of OLS.Among them,station efficiency is the most important factor affecting the passenger flow at rail transit stations,followed by closeness centrality and the ranking of station importance.There are 8 factors contributing more than 5%to passenger flow at stations.Additionally,an independent effect analysis of 9 influencing factors including closeness centrality,station efficiency,and bus stop density reveals nonlinear relationships of these factors with passenger flow at stations and significant threshold effects.The findings provide some technical support for the allocation of Chongqing’s rail transit resources and station planning and offer theoretical reference to the operation of urban rail transit in other cities.