王科伟

入学时间：2014级

答辩时间：2017年

论文题目：信号控制交叉口设计交通量确定方法研究

中文摘要

摘要

城市道路设计交通量是指特定的道路断面，设计年稳定需要通过的交通量与通行能力共同决定了道路设施的功能和性能（服务水平）和规模。然而，以往有关于此的研究却未得到与通行能力相比关于道路设计交通量的研究却未得到应有的重视。所以，深入研究信号控制交叉口设计交通量是科学地进行城市道路规划、设计和管理，并确保其正常运行的前提和基础。

本论文在系统地综述以往相关研究成果的基础上，综合利用交通工程理论、概率论及数理统计和智能算法等方法，针对现有理论与方法在确定设计交通量时存在的局限性，以综合社会成本最佳为目标，亦不仅考虑交通需求，同时，还对信号控制交叉口设计交通量的经济性展开研究。

首先，基于数理统计和假设检验的方法分析了交叉口交通量的分布规律，论证了交叉口进口道小时交通量服从正态分布。其次，根据道路交叉口基础设施成本与出行者的时间成本，以两者成本之和最小对应的交通量定义城市道路交叉口设计交通量。进一步，以交通流的时变规律为基础，确定影响设计交通量大小的社会成本要素，建立了由道路基础设施社会成本与出行者时间成本组成的综合社会成本模型建立综合社会成本模型。

对于新建道路交叉口设计交通量的确定，研究提出了将设计年限的预测年平均日交通量换算为小时交通量，作为交叉口交通量正态分布均值，求得综合社会成本最小为目标的交通量——设计交通量，研究表明，与传统方法相比较可有效地节约社会成本；对于改建道路交叉口，提出了基于工作日全天交通量，以地区周变系数换算到年平均日交通量及小时平均交通量，作为交叉口交通量正态分布均值，以确定设计交通量，与传统方法相比也可明显地节约社会成本。

论文的主要创新点包括：建立了信号控制交叉口设计交通量与社会成本的有机关系；提出了基于社会成本和出行者时间成本最优的新建和改建城市道路信号控制交叉口设计交通量的确定方法。

关键词：信号控制交叉口、设计交通量、综合社会成本、遗传算法

英文摘要

ABSTRACT

Urban road design traffic volume refers to the specific section of the road, the design of the annual stability of the traffic volume and capacity required to determine the road facilities function and performance (service level) and scale. However, the previous research on this aspect has not been compared with the traffic capacity. However, the research on road design and traffic volume has not been paid much attention to. Therefore, the in-depth study of signalized intersection design traffic volume is the premise and basis for scientific urban road planning, design and management, and to ensure its normal operation.

This thesis systematically summarized in the previous relevant research results, the comprehensive utilization of traffic engineering theory, probability theory and mathematical statistics method and intelligent algorithm, aiming at the limitations of the existing theories and methods exist in determining the design of traffic volume, with the goal of optimal social cost, also not only consider the traffic demand, at the same time, also launched study on the economy of the signalized intersection traffic design.

Firstly, based on mathematical statistics and hypothesis testing, the distribution law of traffic volume at intersection is analyzed, and the hourly traffic volume of intersection is assumed to be normal distribution. Secondly, according to the cost of the infrastructure and the time cost of the road intersection, the traffic volume of the urban road intersection is defined according to the minimum cost and the corresponding traffic volume. Further, the traffic flow changes as the basis, to determine the influence of social cost elements design traffic volume, road infrastructure established by the social cost and social cost of establishing comprehensive travel time cost model consisting of the comprehensive social cost model.

To determine the new-bulit road intersection traffic design, the research put forward the design of the annual average daily traffic volume forecast conversion for hourly traffic volume, traffic volume as the mean of normal distribution, to obtain the comprehensive minimal social cost for design traffic volume. The research shows that the model compared with the traditional method effectively reduce the social cost; for the intersection of road reconstruction, proposed traffic volume of working days all day long based on the area with week variable coefficient conversion to the annual average daily traffic volume and average traffic volume, traffic volume as the mean of normal distribution, in order to determine the design traffic volume, compared with the traditional method can be saved obviously social cost.

The main innovations of this paper include: the establishment of the relationship between design traffic volume and social cost of the signal control intersection; put forward a method based on the optimal social cost to determine the design traffic volume of new construction and re-construction signalized intersection.

Key Words:signalized intersection, design traffic volume, total social cost, genetic algorithm

龚越

入学时间：2014级

答辩时间：2017年

论文题目：基于多元数据的交通舆情感知特征提取研究

中文摘要

摘要

互联网已经成为人们生活中不可或缺的一部分，产生了海量的社交数据，包含了大量的交通舆情。对交通运输行业来说，它的点多、面广、服务特色明显，是国计民生关注的重点行业。因此，交通运输行业的舆情感知尤为重要。“交通舆情”是指“在一定的时间空间内，围绕中介性交通事件的发生、发展和变化，作为主体的民众对作为客体的交通要素、交通参与者和交通管理者产生和持有的信念、态度、意见和情绪等表现的总和”。交通舆情感知特征是进行交通舆情的态势感知和安全管理的基础。但是，已有的舆情指标体系普遍存在忽略舆情事件产生具体背景、指向性过宽的问题，不适合交通研究。交通舆情的研究目前也还处于萌芽阶段，面临技术手段研究不足、体系尚未建立的问题。亟需建立完整的特征体系，探索面向应用的技术方法。

本文以交通舆情为研究对象，探讨了交通舆情的概念、作用和指标体系，建立了一套交通舆情采集、分析、挖掘的理论框架体系，对交通舆情的认知和应用，具有重要的指导意义和应用价值。

首先，本文基于我国舆情数据分析报告中的数据，得出了交通舆情信息的三大主要来源：网络新闻、微博和微信。紧接着搭建了基于WebCollector的交通舆情采集框架。提出了交通舆情预处理规则，为后续进一步分析奠定了数据基础。

然后，在总结归纳现有网络舆情指标体系研究成果的基础上，提出了基于舆情的生命周期并面向交通监测与决策建立交通舆情感知特征体系。设立主题特征、时空特征、情感特征、演变特征四个一级特征，交通拥堵、时间、地点等11个二级特征。设计了涵盖舆情数据采集、存储、特征提取到支撑决策的交通舆情感知特征提取应用架构。

随后，对交通舆情感知特征的提取方法进行了探索和研究，建立了一套适用于交通舆情的感知特征提取方法，具有科学性和可行性：

1）提出了基于余弦相似度的主题分类模型。人工标注了3567条文本主题，建立了交通舆情主题分类中文语料库（CCTTC）。检验了文本分类的效果。分类效果符合实际应用要求，但后续仍值得进一步提高。

2） 利用时间表达模式较为固定的特点，建立了基于规则的时间特征提取模型，通过识别文本中的时间表达模式获取时间特征。通过观察上百篇网络文本，建立五层树状空间特征词库，通过识别空间地理词汇提取空间特征。

3）提出了基于词典的交通类文本情感倾向性模型。建立了交通类情感词典，确定了五种情感模式，通过识别情感模式计算文本情感。检验结果显示本模型分类效果出色，符合实际应用要求。

4） 通过分析2016年的“共享单车”网络舆情演变特征，发现交通舆情的波动具有杠杆效应、长记忆性、波动群集性和尖峰后尾性，由此提出了基于EGARCH模型的交通舆情演变特征模型。并提出了对正面舆情和负面舆情分别建模的思想。

最后，本文以深圳公交为研究对象。发现《2017年1季度深圳市公交服务指数》报告中2016年各季度的公交服务指数与本文通过舆情分析得到的公交满意度的趋势一致，证明了本文的研究成果符合实际。通过提取负面舆情关键词，发现“新能源车辆气味重导致晕车”、谣言“新能源车辆辐射大”是导致2017年1季度公交满意度下降的主要原因。同时，提取正面舆情关键词发现，增开线路、提高发车频率、延长公交运营时间等措施更有助于提升乘客满意度。

综上，本文提出的交通舆情感知特征提取方法是可行且有效的。案例分析中也发现，提取得到的舆情数据存在区域分布不均衡、无法细化到具体公交线路上的问题，并不能完全取代传统交通数据的作用。交通舆情更适合宏中观交通研究或交通热点问题研究。本文充分展示了交通舆情与现实交通的相关性以及其中蕴含的巨大价值，结合传统交通数据一定可以在交通研究中发挥显著作用。

关键词：交通舆情，舆情特征体系，网络爬虫，文本分类，时空特征提取，情感倾向性分析，EGARCH模型

英文摘要

ABSTRACT

Nowadays, Internet has become a necessary part of people’s life. Big social data has been produced everyday containing large number of traffic opinions. Transportation is critical to our society and economy, and it’s complex and distinctive, basically it’s a kind of service industry. Therefore understanding users’ suggestions and complaints about transportation is a very significant thing. Traffic opinion is a total of public beliefs, attitudes, opinions and emotions to transportation elements, governments or participants in a certain time space. Big, real time, valuable information is hidden in traffic opinions. Perspective features of traffic opinion are foundation of traffic opinion research, but there are many problems of features and analysis technology. It’s time to build a feature system of traffic opinions and explore the application-orientated technology. This finding can be important for traffic sensing applications where external factors, like sentiment, also contribute on how traffic react.

In this thesis, we aim to automatically extract features from traffic opinions.

Firstly, we propose three sources of traffic opinions referring to Chinese public opinion report. They are netnews, weibo(like twitter) and wechat(like facebook). Then we suggest an approach to collect traffic opinions based on WebCollector, and put forward a rule to preprocess data.

Secondly, we found that present features of opinions are vague and ignore the background and field of opinions. So far, no one has made researches on feature system of traffic opinions. Hence, we propose a feature system of traffic opinions based on life cycle of public opinions for monitoring traffic and making decision. Four Primary characteristics and eleven secondary characteristics are developed in this thesis, in accordance with basic principles of feature building, they are skopos rule, scientfic rule, system rule and application rule. Primary characteristics are theme, time space, sentiment and evolution. Afterwards, we design a framework of how to collect, store, extract and analyze traffic opinions.

Thirdly, we build a set of approaches to extract different features of traffic opinions:

1) We suggest a thematic classification model based on cosine similarity, and build Chinese Corpus of Transportation Topic Classification (CCTTC in short) including 3657 hand-classified texts. Then we test our model with CCTTC data, and the result is acceptable and applicable. But it’s worth improving it in future.

2) We find that many time expressions are repeating. Then we ascertain five modules of time expression, and extract time information by searching time expression module from text. We get space information by searching geographical term from geographical lexicon built by us.

3) We compare rule-based and machine-learning-based methods of sentiment analysis in theory. Therefore, we propose a rule-based sentiment analysis model, in consideration of most traffic terms’ sentiments. We present five sentiment modules and build a sentiment dictionary including positive words, negative words and degree adverbs with their polarities. We calculate text polarity by searching sentiment modules. Then, we test our model and the result is excellent and applicable.

4) We find that the evolution feature of traffic opinions has leverage effect, long memory effect, sheep-flock effect, obvious peak and fat tails. So we propose an EGARCH model and put forward new idea of modeling positive opinions and negative opinions respectively.

Finally, this thesis make an example that how to use approaches which are mentioned above to analyze bus system in Shenzhen, and compare our features extracted from traffic opinion with a report calledBus Service Index during First Season of 2017 in Shenzhen. Then we find that passengers take more care of opening new bus line, increasing departure frequency and extending time of bus service.

In a word, this thesis propose a set of applicable and effective method to extract features of traffic opinions. Athough traffic opinions are vague and its regional distribution is not balanced, and they couldn’t replace traditional traffic data. It’s better to utilize traffic opinions for macro and medium traffic research or hot traffic problem study. Finally, combining traffic opinions with traditional traffic data will make sense on traffic research.

Key Words:traffic opinion, feature system of traffic opinion, web crawler, text classification, extraction of time space feature, sentiment analysis, EGARCH model

张伟丽

入学时间：2014级

答辩时间：2017年

论文题目：城市道路设计车速确定方法研究

中文摘要

摘要

城市土地的高密度开发，用地强度不断增大对交通系统提出了更高的要求。良好的交通系统不仅需要满足快速、高效的要求，更需要其安全、平稳并可靠。城市道路设计车速是指导道路设计与交通优化管理的重要参数，直接关系到设施的几何条件与交通安全和效能及节能减排等。以往的设计车速确定方法多是从运动学角度出发考虑道路几何条件，忽视了道路网络条件及交通条件等的影响。如何平衡道路条件和交通条件对速度的不同要求，是亟待考虑的问题。

本论文拟从系统角度出发，平衡道路几何要素对车速的要求和交通层面对车速的影响。通过对交通系统进行全时空考虑，以交通条件作为切入点对设计车速的确定方法进行再讨论。为了更直观的分析各个影响因素对城市道路车速的影响，给出改建道路设计车速的确定方法，即道路几何条件及路段交通设施确定情况下所对应的设计车速。对于新建道路可根据规划，参考已有经验对各因素数值进行选定。主要开展研究如下：

首先从道路条件、交通条件、广义环境条件分析城市道路设计车速的影响因素，并以道路条件和交通条件作为城市道路设计车速的确定依据，广义环境条件可根据情况进行速度管理。分析了影响设计车速的关键因素如交叉口间距、展宽渐变段、转弯半径、路段出入口、公交停靠站等对车速的影响机理。

其次，分析了城市道路设计车速的制约因素：交叉口间距、展宽渐变段及转弯半径与道路车速的关系，并以此反推设计车速，选取三者中极小值作为设计车速的基础值。

最后，根据设计车速的影响因素对城市道路设计车速基础值进行修正。首先通过分析实地调查数据，解析车道宽度对车速的影响，并给出了车道宽度对车速基础值的修正方法。之后使用微观仿真软件分析了不同出入口间距方差对城市道路车速的影响，并给出在出入口间距方差较大时，设计车速的修正方法。最后，以公交停靠站影响时间表征停靠站对主线车速的影响，并给出公交停靠站对主线设计车速修正的方法。

综上所述，本文从城市道路交通条件功能性及系统性出发，对城市道路设计车速进行再确定。为道路交通设计参数选取提供参考，对后续的研究也有一定的借鉴意义。

关键词：设计车速，车道宽度，出入口间距，公交停靠站

ABSTRACT

The high density of urban land development and increasing land intensity make higher requirements for traffic system. Good traffic system not only needs to meet the requirements of fast and efficient, but also needs its safe, stable and reliable. Urban road design speed is an important parameter for guiding the design of road and optimizing of traffic management, which directly related to the geometry of the road, traffic safety, efficiency, energy saving and so on. Previous determination method of design speed is from the kinematic point of view, while ignoring the impact of traffic conditions. How to balance the different requirements of road conditions and traffic conditions is an urgent issue.

Requirements on speed of road geometric elements and traffic conditions will be balanced from the view of system. Through the whole space-time consideration of the traffic system, the traffic conditions are taken as the key point to discuss the method of determining the design speed. In order to clearly analyze the influence of various factors on the speed, the determination of design speed is based on urban road which geometric condition and traffic conditions are given. For new road, the determination of design speed can be given based on planning, referring to the experience of the various factors selected. The main research is as follows:

Firstly, the influence factors of urban road design speed are analyzed in terms of road conditions, traffic conditions and general environmental conditions. The road conditions and traffic conditions can be used for determining the design speed of urban road and general environmental conditions can give suggestion to speed management. The influence factors of the vehicle speed are analyzed, such as the intersection spacing, the widening of the gradual change section, the turning radius, the access spacing, the bus stop and so on.

Secondly, analyze the relationship between restricting factors and speed: the intersection spacing, the widening of the gradual change section and turning radius. Then the design speed is calculated based on restricting factors. The minimum value of the three factors is chosen as the basic value of design speed.

Finally, the basic value of urban road design speed is modified according to the influence factors. In the first place, the influence of lane width on vehicle speed is analyzed by field survey. After that, the influence of different access spacing on urban road speed is analyzed by using micro simulation software. At last, the influence of bus stop time on the main line speed is analyzed, and the method of modifying the design speed of urban road is given.

In summary, the design speed of urban road is determined by the function and system of urban road traffic conditions, which provides method for the selection of design parameters of road design and gives some references for the subsequent research.

Key words:Design speed, Lane width, Access spacing, Bus stop