交通运输系统工程与信息 ›› 2015, Vol. 15 ›› Issue (2): 189-194.

• 系统工程理论与方法 • 上一篇    下一篇

客票数据“受限”条件下旅客选择偏好估计

骆泳吉,刘军*,马敏书   

  1. 北京交通大学 轨道交通控制与安全国家重点实验室,北京100044
  • 收稿日期:2014-10-31 修回日期:2015-01-13 出版日期:2015-04-25 发布日期:2015-04-27
  • 作者简介:骆泳吉(1988-),男,四川成都人,博士生.
  • 基金资助:

    中央高校基本科研业务费专项资金(2014YJS080);中国铁路总公司科技研究开发计划课题(2013X007-3, 2014X006-A).

Estimating Passenger Choice Preference of Railway Based on Censored Sales Data

LUO Yong-ji, LIU Jun, MAMin-shu   

  1. State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing 100044, China
  • Received:2014-10-31 Revised:2015-01-13 Online:2015-04-25 Published:2015-04-27

摘要:

深入研究客运市场需求,是有效开展客运产品营销工作的前提.区别于既有的以问卷调查为基础的旅客选择行为研究,本文利用客票销售数据研究铁路客运产品选择偏好的估计方法.该方法考虑了部分时段部分列车在供不应求情况下,客票存根数据难以反映旅客真实选择行为的特点,根据预售期内各产品的实时余票数据,对预售期进行划分,建立了分时段的极大似然函数.针对所构建的似然函数,利用EM算法处理流失客流不可获知的问题,从而估计旅客选择偏好.通过实例,估计了徐州东至南京南方向旅客对京沪高铁客运产品的选择偏好.该方法以日常销售数据为基础,能够及时获取旅客选择偏好,具有较强的时效性,可为铁路日常营销工作提供参考.

关键词: 铁路运输, 选择偏好, EM算法, 客票数据, 数据挖掘

Abstract:

A deep study of demand is the basis for passenger transportation marketing. With an increasing attention of passenger choice behavior in recent years, most of the studies estimate passenger preference based on investigation data. In this paper, sales data is utilized to estimate passenger preference. First, a likelihood function considering the availability of products during the sales period is proposed. Since the formulated likelihood is an incomplete-data function, Expectation-Maximization (EM) method is used to solve the incomplete- data problem. An empirical study is taken to analyze the choice preference of highspeed trains for the OD market Xuzoudong to Nanjingnan. As the sales data is convenient to obtain compared to investigation data, this research is meaningful for supporting the daily marketing of railway department.

Key words: railway transportation, choice preference, EM algorithm, sales records, data mining

中图分类号: