张力楠

入学时间：2014级

答辩时间：2021年

论文题目：基于数据的公共交通服务水平评价基础问题研究

中文摘要

摘要

城市公共交通系统是一个综合、复杂、动态的巨系统，对其服务水平进行评价是客观量化公共交通系统服务特征、提高公共交通精细化管理能力，打造高品质城市公共交通系统的关键。这其中，城市地面常规公共汽电车交通系统（以下简称“公交”）在我国公共交通系统中占据主体地位，且对其进行服务水平评价的难度最大、价值最高，研究需求最为迫切。基于数据的公交服务水平评价是面向新环境、新需求来应用新技术的一整套理论与方法体系，论文经过大量查阅相关文献和研究成果，详细解析了公共交通服务水平的评价过程和机理，以及基于数据的概念与内涵，对如何构建客观可比的服务水平评价框架、如何基于典型数据条件挖掘公交服务供给和出行需求特征、如何量化公交服务指标并进行系统应用等基础问题展开了研究，构建了涵盖“多元数据-服务特征-服务评价”的完整公共交通服务水平评价体系。具体有以下几个方面研究内容：

首先，论文以公交为研究对象，从机理解析层面对公交服务系统的服务主体、服务客体和服务过程展开剖析，提出应用弹性分析方法来构建公交服务供给和出行需求之间的匹配关系。结合“基于数据”体系下的相关概念，明确了基于数据的公交服务水平评价在数据输入、评价方法和输出内容上与传统公交服务水平评价之间的差异。之后结合公交乘客满意度调查结果，构建了评价的框架，选取快速性（自由流时间、延误时间）、可靠性（路段通行时间可靠性）、便捷性（换乘时间）、舒适性（拥挤度）四个维度的指标来进行路段级公交服务水平评价和线路级公交服务水平评价。

然后，论文从现实条件出发，明确了我国智能公交系统的典型数据内容。面向服务水平评价需求设计了数据的质量控制流程，并以卫星定位数据和刷卡数据为例，详细介绍了数据质量监测、判断和补全的方法。通过对卫星定位数据进行路段匹配和站点匹配，建立了公交线网和城市道路网络之间的拓扑联系，将传统的公交站间运行特征分析细化到路段乃至十米级的微观层面，实现了公交车辆的运行服务供给特征分析。研究以宁波市慈溪南路为例，运用公交运行速度、通行时间、延误时间和通行时间可靠性等数据结果对公交的路段运行特征进行解耦分析，实现了公交路段运行服务状态和水平的全过程画像，基于数据快速、精准定位公交服务供给短板所在。

论文还提出了公交运行服务供给特征和公交乘客出行需求特征的动态计算框架，通过设置滑动时间窗和计算间隔，不仅确保了每次参与计算的数据样本量，还能够将服务供给和出行需求的特征平滑化，避免由于偶发性因素带来误判。针对公交乘客出行需求特征提取，在通勤模式识别的基础上增加了出行频繁模式挖掘方法，提高了8%的公交个体出行链挖掘率，结合人工跟车调查结果验证了算法的有效性和可靠性。

之后从公交乘客的出行决策过程及心理出发，采用感知行程时间来体现公交服务质量价值并构建了公交路段感知行程时间率模型，以此实现公交服务水平的客观量化。通过分析宁波市2014年11月到2019年11月的公交客流数据和感知行程时间率计算结果，构建了公交服务-客流弹性系数计算模型，计算得到2018年12月的弹性系数为-0.43，对比有关文献中提到的出行时间相关的-0.3至-0.5的弹性范围，该结果为合理数值。

基于弹性系数和感知行程时间率指标构建公交的路段级服务水平和线路级服务水平的评价模型，并将之划分为6个服务水平等级，分别对应20%客流吸引、10%客流吸引、客流吸引力不变、10%客流流失、20%客流流失和30%以上客流流失六种特征。通过赋予公交服务水平等级对应的客流特征，能够帮助公交系统规划建设与运营管理人员选择更合适的预期服务水平标准，以及制定服务水平改善策略。

最后论文面向公交行业主管部门和运营企业的需求，对基于数据的公交服务水平评价与诊断系统的整体架构、具体功能和表现形式进行了设计，并介绍了系统的开发实现过程和系统实例，形成了多个基于数据的公共交通服务水平评价理论与方法的落地应用。

综上所述，本文研究成果对于如何使用典型的公交数据条件实现动态连续、客观可比的公交服务水平评价具有较强的理论指导意义和实践操作价值。

关键词：公共交通，服务水平，数据挖掘，弹性系数

英文摘要

ABSTRACT

Urban public transport system is a comprehensive, complex and dynamic giant system. The evaluation of the level of service is the key to improve the level of refined management and to make the quantitative and objective of the features of public transit service, and it is also the key to build high-quality urban public transport systems. The evaluation of public transit level of service based on data is a set of theory and method system for applying new technology to the new environment and new demand. This paper using a large number of related literatures and research results, starting from the fundamental issues with three levels: level of service framework, indexes and systems. A detailed analysis is used on the process and mechanism of public transit level of service evaluation, and based on the concept and connotation of data, a public transit level of service evaluation flow covering the whole process of "the multivariate data, the characteristics of transit service and the evaluations" is proposed. More researching details are in the following several aspects:

Firstly, from the perspective of mechanism analysis, this paper analyzes the service subject, service object and service process of the level of service. The basic idea of this research is to match the relationships between the supply of public transit service and the passenger flow based on elastic. basic idea of the research is to determine the matching relationship between the supply of bus service and the demand of bus passenger flow through the elastic. Using the concepts from data-based such as data mining, big data and data-driven, then briefly outline the differences between date-based public transit level of service and traditional evaluation methods on data input,methodologyand the output. Then combining with the public transit survey of passenger satisfaction, a framework is built with four dimensions: quickness(travel time of freeflowand travel speed of freeflow) , reliability(link travel time reliability), convenience(transfer time) and comfort(crowded degree), that used for the link level of service and route level of service evaluations.

Then starting from the actual conditions, this paper has found the typical data content of China's intelligent transportation systems, including satellite positioning data, smart card data, fare data, electronic map, line information and station information. And a process of data quality control as conducted for the demand of the evaluation of level of service, an example of data quality detect, judgement and data fill methods using the satellite positioning data and smart card data. When the data preparation is complete, an analysis of the supply of the bus running service, based on the satellite positioning data is taken. Through the study of the link matching and station matching, a topology relationship is built between the bus lines and road network, that can make the analysis of the operation characteristics between bus stations is refined to the micro-analysis level of the road section and even the ten-meter level. Taking the section of Cixi South Road in Ningbo City as an example, the study used the data decoupling analysis of the characteristics of the bus speed track, the travel time and the delay time of each bottleneck to realize the whole process portrait of the service status and level of the bus section, so as to quickly and accurately locate the shortcomings of the bus service level without going out of the house.

This paper also presents a dynamic calculation method for the characteristics of bus operation service supply and bus passengers' travel demand. By setting sliding time window and sliding interval, not only the sample size of each calculation can be ensured, but also the characteristics of operation and demand can be smoothing, so as to avoid misjudgment caused by accidental factors. The Apriori method is used to mining the travel patterns of bus trips, and the mining rate of bus individual trip chain is increased by 8%. The effectiveness and reliability of the algorithm are verified by combining with the results of manual following survey.

Based on the decision-making process and psychology of bus passengers, this paper establishes the model of perceived travel time and the index of perceived travel time rate of bus sections, and calculates it with the spatio-temporal continuous data from the information system of Ningbo Public Transport Corporation. By comparing the five years data in November 2014 to November 2019 of Ningbo, a total of 61 group public transportation passenger flow statistics, ruled out the seasonal factor after the binomial regression equations are established and the calculation of passenger flow in December 2018 perception rate of travel time elasticity coefficient E = 0.43. Compared with the travel time mentioned in the literature related elastic range 0.3 to 0.5, the result is reasonable.

Then constructed by elastic coefficient and perception rate of travel time is given priority to public transit bus link level of service and route level of service evaluation formula. It could be divided into six service levels level, corresponding to 20% to attract, 10% attract passenger flow and traffic passenger flow remains the same, 10%, 20% loss of passenger flow and traffic loss more than 30% of the six characteristics of passenger flow erosion by giving service level of clear service utility meaning, can assist the public transportation system planning and construction and operation management personnel more appropriate choices expected service level standard, and to develop service level improvement strategy.

Finally, this paper designed the system architecture, specific functions and manifestations of the data-based bus service level evaluation and diagnosis system, and introduces the development and implementation process of the system and application examples, forming a closed loop of data analysis and application.

In conclusion, the research results of this paper have strong theoretical guiding significance and practical value for how to use typical bus data conditions to achieve dynamic, continuous and objectively comparable public transit level of service evaluation.

Key Words:Public Transit, Level of Service, Data Mining, Elasticity