【作者】 贺琳；

【导师】 陈燕；

【作者基本信息】 大连海事大学 ， 管理科学与工程， 2012， 博士

【摘要】 道路运行安全风险分析是道路交通安全管理的一个较新的研究方向。目前的分析方法,大都从管理制度上进行论述,或者采用基本的统计方法进行宏观层面的规律性分析,缺少先进的、智能化的分析方法。自2006年8月交通部发布《全国公路交通阻断信息报送制度》以来,交通部及各省级路网中心累积了大量的区域路网交通阻断信息,引入“数据挖掘”理论与技术,充分利用历史道路运行数据,梳理和分析干扰道路安全运行的各项原因,发现道路安全运行管理的内在规律,为我国道路安全运行制定科学的安全防范措施,提供辅助决策支持,对提高我国道路交通安全管控能力具有十分重要的现实意义。本文在分析国内外数据挖掘技术、交通安全理论和两者相结合方法的研究及其应用现状基础上,以国家科技支撑项目“国家高速公路安全和服务技术开发与工程应用示范”实践为背景,应用多学科知识交叉融合来研究基于数据挖掘的道路运行安全风险分析方法。首先,论文研究了道路运行安全风险相关信息的数据来源、数据特点、以及各个组成要素,进而提出了一个基于本体的道路运行安全风险数据仓库建模方法,该方法按照主题-维度-类别的方式,通过扩展和新增的BWW本体构件,将用户需求划分为多个主题,每一个主题由多个维度刻画,每个维度细分为多种类别,并进行维度、类别等对象的特性及其之间的关系的形式化定义和可视化描述。通过该方法,建立具有语义信息的概念模型,完成基于数据仓库的多源信息整合任务,为道路运行安全风险分析和管理提供有效的数据整合技术保障。然后,以“风险因素”分析为重点,在归纳和总结诱发道路交通阻断原因基础上,提出了采用改进的风险指数评价法来完成道路运行安全风险因素评估,通过对概率和严重性的动态划分获得各因素的相对风险指数,实现了对区域路网各动态因素风险性的量化评估,进而构建了道路运行安全评价指标体系,完成道路运行安全的综合评价。其次,以“风险事件”研究为重点,分析阻断事件的时间和地域等分布情况,提出了基于模糊聚类的道路交通阻断等级划分方法和基于模糊关联规则的公路运行安全成因分析方法,上述方法根据历史实际数据,实现了道路运行安全阻断等级的准确划分,并挖掘出交通事件属性之间的依存关系,为交通阻断的预防和处置提供辅助支持。再次,以“异常状态预测”研究为重点,针对现行道路运行异常状态的获取方式和处理模式效率低的问题,提出了一种道路运行异常状态预测方法,基于组合预测的路段交通量预测分析模型和基于RBF断面交通流参数(流量、速度、占有率)的预测偏差分析模型,实现对道路运行的常发性异常状态和偶发性异常状态进行动态预测。最后,将论文研究成果与工程实际结合,以某道路主管部门的风险分析需求、应用背景和实绩数据,来验证本文所述方法的有效性和可行性,研究成果为道路运行安全风险智能化和知识化管理提供了一定的参考与借鉴。更多还原

【Abstract】 The risk analysis for highway safety operation is a new study area of highway traffic safety management. The current analysis methods mostly discuss the management system or analyze regularity of the regularity in the macroscopic level using the statistic method, lacking the advanced intelligent quantitative method. Since August 2006 the submitting system of national highway traffic blocking information issued by ministry of transport of the people’s republic of China, each provincial highway network center have accumulated a lot of regional network traffic blocking information. By data mining theory and technology how to use history highway operation data, analyze the highway safety operation of the interference various reasons and find the inner rules of highway safety operation management, can provide decision support for making a scientific safety countermeasures of our country highway safety operation, which is important practical significance to enhance our country’s highway traffic safety control ability.On the basis of analyzing the domestic and overseas research and application status of data mining technology, traffic safety theory and its combined method, as well as under the background of practice and application in the national science and technology support project of application demonstration of national highway safety and service technology development and engineering, a number of key technologies of highway safety operation risk analysis method based on data mining are studied by this paper through using multi-disciplinary knowledge and cross-application integration.Firstly, the paper studies the multi- source information integration method for highway safety operation risk management based on data mining, and introduces data source, data characteristics and composing elements, and puts forward a risk data warehouse conceptual modeling method of highway operation based on ontology, according to the subject-dimensions-category way which divided the user demand into multiple topics, each topic described by several dimensions characterizations and each dimension divided into a variety of categories by expanding the new BWW component ontology. Through the method we can make formal definitions and visual descriptions for the relationships among the topics, dimensions and categories in order to accomplish specific highway operation safety risk analysis.secondly, with risk factors as the research core, the paper proposes a highway safety operation risk factor assessment method using improved risk index evaluation method based on Summing up the highway traffic block reasons, and builds up a highway safety operation risk evaluation index system with dynamic classification of the probability and the severity, in order to realize he quantitative evaluations of the regional network dynamic risk factors and the comprehensive risk assessment of highway operation safety. Then, with risk events as the research core, the paper proposes a highway traffic block hierarchy method based on fuzzy clustering and a risk reasons analysis method for highway safety operation based on fuzzy association rules on the basis of mining the distribution of the highway blocking event, the highway blocking time and the highway blocking area, which realize the accurate division of highway safety operation blocking levels and mining the traffic events interdependence relationships in order to provide auxiliary support for traffic block prevention and treatment. Moreover, with highway operation state prediction as the research core, In view of the low efficiency of the current abnormal highway state access and treatment, the paper proposes a highway operation abnormal state prediction method based on a prediction deviation analysis model,which including road traffic flow combination forecasting and section traffic flow parameters (flow, speed, road share) forecasting through RBF, in order to predict dynamically the frequently abnormalities state and accidental abnormalities state in the highway operation process. Finally, by combining the thesis research and real-world practice, and following the case a highway authority requirements of risk analysis, application background as well as performance data, the results have demonstrated the effectiveness and practicability of the cost management method based on data mining. The research fruits can be cited and referred by other traffic operation safety management for achieving their intelligent and knowledge-based risk management.更多还原