基于时空相关性分析的交通流预测研究

发布时间：2022-11-05 09:41

　　机动车数量的快速增加以及城市化进程的扩展加剧了不断增长的交通需求与城市交通基础设施供应能力之间的矛盾。作为城市交通面临的巨大压力导致的社会问题之一,交通拥堵已经成为在交通管理领域中急需解决的严重挑战。能够缓解交通拥堵的最可行的措施是通过建立智能交通系统（Intelligent Transportation System,ITS）,以提高交通管理和服务效率。智能交通系统由一系列能够为交通管理者、车辆以及个别出行者提供多种服务的先进技术所组成,从而使交通系统的各个部分能够更好的协调,共享有用信息,做出及时正确的决策。交通流预测是在智能交通系统研究和应用中不可缺少的重要功能,如何实时准确地预测未来交通流已经成为交通管理科学领域中的一个研究热点。许多研究人员提出了多种交通流预测模型。然而,关于如何挖掘影响目标交通流的不同因素以及如何将此信息结合在预测模型中的研究仍然不足。为了提高交通流预测模型的性能,必须考虑与目标交通流有关的足够信息。在交通网络中相邻路段上的交通流之间存在明显的时空相关性。因此,正确估计此类相关性对提高交通流预测结果的准确性至关重要。针对交通流时空相关的非线性以及动态性分析,... 

【文章页数】：158 页

【学位级别】：博士

【文章目录】：
Abstract
摘要
Chapter 1 Introduction
    1.1 Research Background
    1.2 Objectives and Significances of the Research
    1.3 Parameters and Characteristics of Traffic Flow
        1.3.1 Traffic Flow Parameters
        1.3.2 Characteristics of Traffic Flow
    1.4 Overview of Related Works
        1.4.1 Traffic Flow Prediction Models
        1.4.2 Spatiotemporal Correlation Analysis
        1.4.3 Traffic Pattern Clustering
    1.5 Main Contents of the Research
    1.6 Structure of the Dissertation
Chapter 2 Nonlinear Analysis of Spatiotemporal Correlation Based on Mutual Information
    2.1 Mutual Information
    2.2 MI-Based Feature Selection Criterion and Search Strategy
    2.3 Predictor Selection Using MI-Based Feature Selection Criterion
        2.3.1 Generation of Features from Traffic Time Series
        2.3.2 Feature Selection Algorithm
        2.3.3 Traffic State Vector
    2.4 Adaption of KNN-Based Prediction Model
        2.4.1 KNN-Based Prediction
        2.4.2 Composition of Traffic State Vector
        2.4.3 Distance Metric and Prediction Function
    2.5 Case Study
        2.5.1 Case Data and Evaluation Measures
        2.5.2 MI Estimation Method
        2.5.3 Results of MI Analysis
        2.5.4 Feature Selection and Traffic State Vector Composition
        2.5.5 Prediction Results
    2.6 Brief Summary
Chapter 3 Traffic Clustering for Dynamic Analysis of Spatiotemporal Correlation
    3.1 Clustering of Spatiotemporal Correlation Matrices
        3.1.1 Spatiotemporal Correlation Matrix
        3.1.2 Traffic Clustering Based on CLARANS
    3.2 Spatiotemporal Correlation Analysis
    3.3 Cluster-Wise Predictor Selection
    3.4 Case Study
        3.4.1 Experimental Settings
        3.4.2 Clustering Results
        3.4.3 Results of Spatiotemporal Correlation Analysis
        3.4.4 Results of Predictor Selection
    3.5 Brief Summary
Chapter 4 Prediction-After-Classification of Traffic Flow
    4.1 Overall Framework
    4.2 Offline Phase
    4.3 Online Phase
        4.3.1 Classification of Current Traffic Pattern
        4.3.2 Regime-Switching Prediction
    4.4 Case Study
        4.4.1 Comparative Prediction Models
        4.4.2 Prediction Results
        4.4.3 Comparison with Other Clustering Methods
    4.5 Brief Summary
Chapter 5 Some Application Models for Traffic Management
    5.1 Mining of Spatiotemporal Correlation in Entire Traffic Network
    5.2 Traffic Flow Prediction over Entire Traffic Network
    5.3 Network Decomposition for Distributed Traffic Management
    5.4 Dynamic Route Guidance with Predictive Traffic Information
        5.4.1 Route Guidance System
        5.4.2 Routing Strategy with Predictive Traffic Information
        5.4.3 Simulation
    5.5 Brief Summary
Conclusion
结论
References
List of Publications During the Doctoral Degree
Acknowledgement
Resume


【参考文献】：
期刊论文
[1]A two-stage short-term traffic flow prediction method based on AVL and AKNN techniques[J]. 孟梦,邵春福,黃育兆,王博彬,李慧轩.  Journal of Central South University. 2015(02)



本文编号：3702370