基于分类学习的铁路客票销售预警技术研究

发布时间：2019-01-09 13:35

【摘要】：与民航运输相比,铁路客运中不同OD旅客之间在运输能力上的竞争较为激烈,尤其是对我国铁路客运而言。由于我国列车开行距离总体较长且运能紧张的问题尚未彻底解决,因而这种矛盾尤为突出。为解决这一矛盾,我国铁路客运部门逐渐形成了一套围绕票额管理的售票组织策略,用于调节列车运能对沿途不同OD旅客需求的满足程度,其基本原则是:优先满足长途客流,在此基础上,尽量公平地满足短途客流。然而,在预售期开始之前的客流预测难以达到理想的精度,故需要在预售期内根据售票情况对有可能存在不合理策略的列车进行预警,以及时调整售票组织策略。本文正是围绕这一问题,试图提出一种基于分类算法的铁路客票销售预警方法,以克服目前主要依靠管理人员人工判断的效率低下,不能适应近年来列车数量剧增且运营管理需求日益精细化要求的不足。本文首先对客票销售中的售票组织动态调整及预警业务进行了详细分析,明确了预售期内客票销售预警的业务需求。通过对铁路客运营销辅助决策系统中列车历史运营数据的分析,提出了用于描述列车特征及客票销售过程特征的属性变量,并总结出了五种可能的销售结果类型。在此基础上,结合贝叶斯分类方法在不确定性推理方面的优势,构建了基于贝叶斯分类的客票销售预警模型,利用该模型可对旅客列车客票销售进程进行评估和预警。论文最后选取了京沪高铁列车历史运营数据,通过基于实际数据的仿真实验验证了所提出的预警模型及算法的可行性,并对其预警结果准确度与列车开行日期类型、距离发车时间、客座兑现率等因素的关系进行了分析。
[Abstract]:Compared with civil aviation transport, the competition between different OD passengers in railway passenger transport is more fierce, especially for China's railway passenger transport. Because the train running distance is long in our country and the problem of shortage of transportation capacity has not been solved thoroughly, this contradiction is especially prominent. In order to solve this contradiction, the railway passenger transport department of our country has gradually formed a set of ticketing organization strategy around ticket quota management, which is used to adjust the degree of satisfaction of train capacity to different OD passengers along the way. The basic principle is: first to satisfy the long-distance passenger flow. On this basis, as fair as possible to meet the short distance passenger flow. However, the forecast of passenger flow before the pre-sale period is difficult to achieve the ideal precision. Therefore, it is necessary to warn the trains with unreasonable strategies according to the ticket sales situation during the presale period, in order to adjust the ticketing organization strategy in time. This paper attempts to propose an early warning method of railway ticket sales based on classification algorithm to overcome the inefficiency of mainly relying on manual judgment of managers at present. It can not adapt to the shortage of train quantity increasing sharply and operation management demand becoming more and more fine in recent years. In this paper, the dynamic adjustment of ticket sales organization and the early warning business are analyzed in detail, and the business requirements of ticket sales early warning during the pre-sale period are clarified. Based on the analysis of the historical operation data of the railway passenger transport marketing assistant decision system, the attribute variables used to describe the train characteristics and the characteristics of the passenger ticket sales process are put forward, and five possible sales result types are summarized. On this basis, combined with the advantage of Bayesian classification in uncertainty reasoning, a ticket sales early warning model based on Bayesian classification is constructed, which can be used to evaluate and warn passenger train ticket sales process. Finally, the paper selects the historical operation data of Beijing-Shanghai high-speed train, and verifies the feasibility of the proposed early warning model and algorithm through the simulation experiment based on the actual data, and gives the accuracy of the warning results and the type of the train departure date. The relationship between distance departure time, guest rate and other factors is analyzed.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U293.22

【参考文献】