基于动态张量填充的短时交通流预测研究

发布时间：2018-02-05 05:31

本文关键词： 短时交通流预测动态张量填充多模式分析矩阵分解　出处：《北京理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：智能交通系统ITS（Intelligent Transportation Systems）是近年来发展起来的交通控制管理信息系统。作为智能交通系统重要组成部分的交通控制系统，交通管理系统以及交通诱导系统都要求为其提供准确的实时交通信息，而在实时信息基础上的短时交通流预测则是实时控制和诱导的前提，是智能交通系统的重要理论基础。交通流数据中蕴含着丰富的多模式特征，如何在统一框架下充分利用交通流数据的多模式特性仍然是一个难点问题，鉴于此，本文针对短时交通流预测问题，利用张量多线性模型在多模式上动态地对交通流数据进行分析，，构建动态交通张量模型，研究动态张量模型框架下的短时交通流预测方法，本文主要内容具体包括以下几个方面：第一，从多模式信息挖掘角度出发，分别构建交通流数据静态张量模型以及动态张量模型，在张量框架下利用多线性分析对交通流数据进行分析，揭示交通流数据的多模式低秩特征，提出以交通流数据的多低秩性为基础的短时交通流预测问题。第二，在交通数据动态张量模型基础以及多模式分析基础上，提出基于矩阵分解的张量填充方法解决短时交通流预测问题，并由此提出一种以多模式矩阵分解为基础的张量填充方法及其理论。并从多个方面测试基于矩阵分解的张量填充方法的可行性。最后，结合动态张量填充方法与交通流动态张量模型，提出基于动态张量模型的短时交通流预测方法，并与现有的预测方法进行对比实验，分别在正常数据下和丢失数据下对算法进行检验测试，实验结果表明，在一定条件下本文提出的方法能够准确地预测交通流量且能在统一框架下准确地填充丢失交通数据以及预测未来交通流量。
[Abstract]:Intelligent transportation system ITS (Intelligent Transportation Systems) is a traffic control information management system developed in recent years. As an important part of the intelligent traffic control system, traffic system, traffic management system, traffic guidance systems are required to provide accurate information for the real-time traffic, and traffic information in real time based on flow prediction is the real-time control and guidance of the premise, is an important theoretical basis for the intelligent traffic system. The traffic flow data contained in the multi mode feature rich, how to make full use of the multi mode characteristics of traffic flow data in a unified framework is still a difficult problem, in view of this, based on the short-term traffic flow prediction problem, using multi linear tensor in the multi model dynamic model of traffic flow data analysis, constructs a dynamic traffic tensor model, study on dynamic tensor model under the framework of the Short time traffic flow forecasting method, the main contents of this paper include the following aspects:
First, starting from the perspective of multimodal information mining, constructs the traffic flow data of static model and dynamic model of tensor tensor in the tensor framework, using linear analysis to analyze the traffic flow data reveal multi mode low rank characteristic data of traffic flow, the low rank of the traffic flow data in short-term traffic based flow prediction problem.
Second, based on the dynamic traffic data tensor model and the multi mode on the basis of the analysis, put forward the method to solve the tensor matrix decomposition with short term traffic flow forecasting based on the problems and put forward a kind of tensor based on multi pattern matrix decomposition based filling method and theory. And from the aspects of feasibility test based on tensor matrix decomposition filling method.
Finally, combined with the dynamic tensor filling method and dynamic traffic flow tensor model, proposed a prediction method of short term traffic flow based on dynamic tensor model, and is compared with the existing prediction methods, missing data under test to test the algorithm respectively in normal data and experimental results show that under certain conditions, this can accurate prediction of traffic flow and can accurately fill in a unified framework of lost traffic data and to predict the future traffic.

【学位授予单位】：北京理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.14

【参考文献】