摘要
科学的轨道交通出行模式分析是运营决策优化的重要依据。为挖掘城市轨道交通时空流动特征及其影响机理,提出一种基于非负张量分解的OD客流强度时空分布计算方法,采用融合SHAP归因分析的极端梯度提升树(eXtreme Gradient Boosting,XGBoost)对各模式OD客流强度进行拟合预测。使用城市轨道交通AFC(automatic fare collection system,AFC)系统数据,从空间、时段以及出行日3个维度构建3阶客流OD张量,采用交替非负最小二乘法(alternating non negative least squares,ANLS)实现非负CP张量分解。基于张量分解结果,从北京轨道交通344个站点连续1周16266966条出行数据中,提取出晨高峰长距离通勤、早高峰中短通勤、平峰休闲中转出行、晚归出行4种出行模式的时、空分布特征。基于可解释性机器学习模型,对各模式OD客流进行预测。结果表明XGBoost与CatBoost、LightGBM、OLS相比更具优势。根据OD起终点站域环境特征,考虑起终点缓冲区内各类兴趣点(point of interest,POI)数量、小区住户数、房价、人口数量、站点偏离距离以及出行距离等指标,构建OD强度关联指标体系,解释各指标对OD客流强度的正负反馈效应。SHAP归因分析说明,居民更倾向于14站以内的中短途出行,并分别得到了就业类POI数目对晨、早通勤客流正向影响,以及餐饮类POI数目对休闲中转出行客流正向影响的临界阈值。该方法可为轨道交通精细化出行引导和客流组织提供数据支撑,优化城市轨道交通供需平衡及服务水平。
A scientific analysis of the urban rail transit travel mode is an important basis for the optimization of operational decision-making.To understand the spatial and temporal flow characteristics of urban rail transit and its mechanism of influence,a method based on non-negative tensor decomposition was proposed to calculate the time and space distribution of OD ridership.The ridership intensity of each OD pair was fitted and predicted by eXtreme gradient boosting tree model(XGBoost)with SHAP attribution analysis.Using the Automatic Fare Collection System(AFC)data of urban rail transit system,a three-dimensional OD tensor of passenger flow was constructed from the degrees of spatial,time and date of travel.The non-negative CP tensor decomposition was realized by alternating non-negative least squares(ANLS)method.Based on the tensor decomposition,the time and space distribution features of four travel modes,which were the long-distance commuting during early morning rush hour,short and medium commuting during morning rush hour,leisure transit and transfer during off peak.The late return travel were generated from 16266966 trip data on 344 urban rail stations in Beijing.The OD intensity for each travel mode was predicted by the explainable machine learning model.It can be concluded that XGBoost has more advantages compared with CatBoost,LightGBM,and OLS.Considering the environmental characteristics of OD terminal stations,the correlation index system of OD intensity is constructed to explain the positive and negative feedback effects of each index on the OD passenger flow intensity,including the number of various points of interest(POI)in the buffer zone of the stations of OD pairs,the number of households in the community,the housing price,the population,the departure distance from the center and the travel distance.SHAP attribution analysis shows that residents are more inclined to travel short to medium distances within the 14 stations.The critical thresholds for the positive impact of the number of employment POIs on morning commuting passenger flow and that of the number of catering POIs on leisure transit passenger flow were obtained.This method can provide data support for the elaborate travel guidance and passenger flow organization to optimize the supply-demand balance and service level of urban rail transit.
作者
周雨阳
李世堃
胡世龙
邓沙沙
柳堤
陈艳艳
罗铭
ZHOU Yuyang;LI Shikun;HU Shilong;DENG Shasha;LIU Di;CHEN Yanyan;LUO Ming(Beijing Key Laboratory of Traffic Engineering,Beijing University of Technology,Beijing 100124,China;Key Laboratory of Advanced Public Transportation Sciences,Ministry of Transport,P.R.China,Beijing University of Technology,Beijing 100124,China;Beijing Engineering Research Center of Urban Transport Operation Guarantee,Beijing University of Technology,Beijing 100124,China;Traffic Control Technology Co.,Ltd.,Beijing 100070,China)
出处
《铁道科学与工程学报》
北大核心
2025年第5期2000-2012,共13页
Journal of Railway Science and Engineering
基金
北京市自然科学基金−丰台轨道交通前沿研究联合基金资助项目(L231025)
教育部人文社会科学研究规划基金资助项目(23YJAZH228)。
关键词
城市轨道
出行模式
张量分解
机器学习
归因分析
urban rail
travel mode
tensor decomposition
machine learning
attribution analysis
