基于案例推理的高速公路清障救援资源需求预测研究

发布时间：2018-05-30 09:20

本文选题：高速公路 + 清障救援　；参考：《交通运输部公路科学研究所》2017年硕士论文

【摘要】：随着我国高速公路里程与机动车保有量的迅速增长,使得高速公路的安全、高效运营的保障任务日益艰巨。作为高速公路“保通保畅”工作实施的重要基础条件,清障救援资源对减少高速公路交通事故所造成生命财产损失、提高路网通行效率等方面发挥着重要作用。针对清障救援资源的配置管理使用,合理地预测需求是进行后续资源配置调度等工作的前提条件。研究高速公路清障救援资源配置需求预测问题具有重要的理论与实际意义。通过介绍案例推理方法理论框架并分析其应用于应急救援领域的优势,将案例推理引入清障救援资源需求预测。通过分析常用知识表达方法,提出资源预测案例的二元组框架表示方法。从道路路线、路基路面等方面分析高速公路交通事故影响因素,构建包含道路长度等要素在内的案例特征属性空间。根据目前救援资源实际使用状况及装备参数统计数据,形成案例数量结果的分类型分能力表示方法。针对案例特征属性空间构建过程中可能存在的冗余信息,依据特征选择过程一般过程及搜索策略,在阐述随机森林方法原理、构建过程、关键参数、应用优势的基础上,应用随机森林方法袋外误差降低率衡量特征重要性程度,采取后向搜索策略对案例特征进行选择。基于实际数据给出了特征算法的实例应用过程,确定包含了道路长度,交通量等关键因素的清障救援车辆数量预测特征集合。通过对传统案例检索方法的优缺点分析,提出结合径向基神经网络的学习案例检索以提高检索过程效率。通过介绍径向基神经网络一般原理及常用学习方法,针对径向基网络隐含层难以确定的问题,设计了利用动态衰减方法改进径向基网络学习方法,并经过实例数据验证改进后的径向基神经网络检索能在保证检索精度的情况下有效提高学习的速度。
[Abstract]:With the rapid growth of highway mileage and the number of motor vehicles in our country, the task of ensuring highway safety and efficient operation is becoming increasingly arduous. As an important basic condition for the implementation of the expressway "Baotong Baochang", the obstacle clearing and rescue resources play an important role in reducing the loss of life and property caused by the expressway traffic accident and improving the efficiency of the road network. In view of the allocation management and use of obstacle clearing and rescue resources, reasonable prediction of demand is the prerequisite for the subsequent resource allocation and scheduling. It is of great theoretical and practical significance to study the demand prediction of freeway rescue resource allocation. By introducing the framework of Case-Based reasoning (CBR) and analyzing its advantages in the field of emergency rescue, Case-Based reasoning (CBR) is introduced to predict the resource demand of obstacle clearing and rescue. Based on the analysis of common knowledge representation methods, a binary group framework representation method for resource prediction cases is proposed. This paper analyzes the influence factors of expressway traffic accidents from the aspects of road route, roadbed and pavement, and constructs the case characteristic attribute space including road length and other factors. According to the statistical data of the actual use of rescue resources and the equipment parameters at present, a method of expressing the ability of classification and classification of the results of the number of cases is formed. According to the general process of feature selection and the search strategy, the principle, construction process, key parameters and application advantages of stochastic forest method are expounded, based on the redundant information that may exist in the process of constructing case feature attribute space, based on the general process of feature selection and the search strategy. The importance of feature is measured by the reduction rate of out-of-bag error of stochastic forest method, and the case feature is selected by backward search strategy. Based on the actual data, the application process of the feature algorithm is given, and the prediction feature set of the number of obstacle clearing and rescue vehicles including road length, traffic volume and other key factors is determined. Based on the analysis of the advantages and disadvantages of the traditional case retrieval methods, a learning case retrieval method combined with radial basis function neural network (RBF) was proposed to improve the efficiency of the retrieval process. This paper introduces the general principle and common learning methods of radial basis function neural network, aiming at the problem that the hidden layer of radial basis function network is difficult to be determined, a dynamic attenuation method is designed to improve the learning method of radial basis function network. The improved radial basis function neural network retrieval method can effectively improve the learning speed under the condition of guaranteeing the retrieval accuracy.
【学位授予单位】：交通运输部公路科学研究所
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U492.8

【参考文献】