操春燕
随着社会的发展，城市交通问题越来越严重，也越来越受到人们的重视，所以对城市路网的有效管理与控制也显得越来越重要。交叉口是路网的节点，对交叉口的有效管理和控制是改善整个路网交通的关键。对交叉口进行管理和控制时，交通流量是不可或缺的基本数据，对于那些有检测器的交叉口可以直接获取流量数据，对于那些无检测器的交叉口就需要进行预测，因此，交叉口交通流量预测值得深入研究。国内外在这方面也做了大量的研究，本论文正是在之前研究成果的基础上，提出了自己的研究思路和成果。总体来说，本论文主要有以下几方面内容：
（1）绪论。简要介绍了本论文的研究背景、现状、内容、目的及意义。
（2）路网内交叉口相关性分析。本章通过实例对比分析了三种交叉口相关性的研究方法，并为后面的交通流量预测提供了基础。
（3）交叉口交通流量预测模型概述。通过对比之前的转弯流量推算和交通流量预测方法，提出了利用BP神经网络的方式，基于路网中有检测器交叉口交通流量，来预测路网中无检测器交叉口的转弯流量。
（4）BP神经网络模型的建立。根据第3章提出的研究思路，建立可行的BP神经网络预测模型。
（5）对BP神经网络模型的训练和仿真。利用matlab工具，应用第2章的实例和结论，对第4章建立的BP神经网络模型进行训练和仿真。
（6）本研究的结论和展望。
通过实际数据对本论文所提出的模型进行了检测，并对今后的研究提出了建议和意见。
关键词：交叉口相关性，聚类分析，主成份分析，逐步回归分析，交通流量预测，BP神经网络，matlab
With rapid development of society, the problem caused by traffic is becoming more and more serious. And people also pay more attention to it. So it is important that effectively managing and controlling the traffic of the city network. Intersection management is the key to improve traffic condition of the whole network. Because traffic volume is the basic data, we need to obtain it. For those having the detector intersection, we can directly get it, but having no detector intersection need to estimate it. Therefore traffic volume forecasting is very important. There have been much research in foreign and domestic, but this article put forward own ideal and result, as follows:
(1) At first, it was the summary of the whole paper, including the research background, present condition, contents, purpose and meaning of this thesis.
(2) The second chapter analyzed three kinds of research methods on intersection relativity by an example. And the result was the foundation for traffic volume forecasting.
(3) In third chapter, the paper introduced some traffic volume forecasting model in brief.
(4). According to chapter 3’s research, the fourth chapter established a BP neural network model for traffic volume forecasting.
(5) The fifth chapter trained and simulated the BP neural network model.
(6) Final, it was the conclusion of the paper.
Based on the research result, the paper also put forward some suggestion and opinion to the research of aftertime.
Key words: intersection relativity, cluster analysis, principal component analysis, step-by-step regression analysis, traffic volume forecasting, BP neural network.
