首页|高速铁路预售期旅客购票量分布预测

高速铁路预售期旅客购票量分布预测

扫码查看
原文链接
  • 万方数据
  • 维普
在预售前(相隔31 d)预测高速铁路预售期旅客购票量分布是铁路企业精准进行收益管理的前提.基于高速铁路预售模式和旅客售票数据,分析预售期内各预售日旅客购票量的相关性,探究预售期旅客购票量分布的影响因素.综合考虑出发日特征以及旅客购票量分布时序特征的影响,构建了考虑多输出间关联性的最小二乘支持向量回归-卷积长短期记忆网络(MLSSVR-ConvLSTM)模型.以京沪高铁线路中上海虹桥站至北京南站、上海虹桥站至徐州东站、上海虹桥站至无锡东站这3种不同距离OD旅客为例,进行预售期旅客购票量分布预测实例分析.研究结果显示:MLSSVR-ConvLSTM模型预测结果较好地反映了真实的预售期旅客购票量分布的变化趋势,平均绝对百分比误差为6.7%～11.0%,预测效果优于多元线性回归(MLR)、K近邻回归(KN)、极致梯度提升算法(XGBoost)、支持向量回归机(SVM)、多输出最小二乘支持向量回归(MLSSVR)和卷积长短期记忆网络(ConvLSTM)等模型,验证了所提出模型的合理性和有效性.进而表明,在构建预售期旅客购票量分布预测模型时,考虑预售期旅客购票量分布整体性以及各类因素的综合影响可有效地提高模型预测精度.所提出的预售期旅客购票量分布预测模型可以为铁路企业制定动态票额分配和浮动票价等政策提供理论支撑.
Prediction of passenger booking volume distribution during pre-sale period for high-speed railways
Predicting the passenger booking volume distribution during pre-sale period for high-speed railways before pre-sale(31 days apart)is the premise for accurate revenue management by railway enterprises.Based on the pre-sale mode of HSR and passenger ticket sales records,this paper analyzed the correlation of booking volume on each pre-sale day during the pre-sale period,and explored the influencing factors of passenger booking volume distribution during pre-sale period.A multi-output least squares support vector regression-convolutional long short-term memory network(MLSSVR-ConvLSTM)model considering the correlation between multiple outputs was constructed,taking into account the departure date characteristics and the temporal characteristics of passenger booking volume distribution.Taking the OD passengers under three different distances from Shanghai Hongqiao Station to Beijing South Station,Shanghai Hongqiao Station to Xuzhou East Station,and Shanghai Hongqiao Station to Wuxi East Station in the Beijing-Shanghai HSR line as examples,the empirical analysis of the prediction of passenger booking volume distribution during pre-sale period was conducted.The results show that MLSSVR-ConvLSTM model prediction results can reflect the change trend of the real passenger booking volume distribution during pre-sale period,with the mean absolute percentage error ranging from 6.7%to 11.0%.The prediction effect is better than multiple linear regression,K-nearest neighbor regression,extreme gradient boosting,support vector regression machine,multiple output least squares support vector regression,and convolutional long short-term memory network models,which verifies the reasonableness and validity of the proposed model.It further shows that when constructing the prediction model for passenger booking volume distribution during pre-sale period,considering the integrity of passenger booking volume distribution during pre-sale period and the comprehensive influence of various factors can effectively improve the prediction accuracy of the model.The proposed model for prediction of passenger booking volume distribution during pre-sale period can provide theoretical support for railway enterprises to formulate policies such as dynamic ticket allocation and floating fares.

high-speed railwayspre-sale periodprediction of passenger booking volume distributionMLSSVR-ConvLSTM modelticket sales records

徐光明、林珊珊、米希伟、王凯、胡心磊

展开 >

中南大学 交通运输工程学院,湖南 长沙 410075

北京交通大学 交通运输学院,北京 100044

高速铁路 预售期 旅客购票量分布预测 MLSSVR-ConvLSTM模型 售票数据

国家自然科学基金资助项目国家重点研发计划资助项目湖南省自然科学基金资助项目京沪高速铁路股份有限公司科技研究项目

721712362020YFB16004002022JJ30767京沪科研-2022-1

2024

10.19713/j.cnki.43-1423/u.T20230293

铁道科学与工程学报
中南大学 中国铁道学会

铁道科学与工程学报

CSTPCD北大核心EI
影响因子：0.837
ISSN：1672-7029
年,卷(期)：2024.21(1)
  • 9