基于支持向量回归机和多变量相空间重构的短时交通流预测
发布时间:2018-01-25 22:17
本文关键词: 多变量 混沌 相空间重构 支持向量机 交通流预测 出处:《重庆交通大学》2014年硕士论文 论文类型:学位论文
【摘要】:实时准确的交通流预测是交通信号控制系统和交通流诱导系统应用的前提和关键,其预测精度直接关系到交通控制和交通诱导的运行效果。由于交通系统具有随机性、时变性、强非线性等特点,因此人工智能方法越来越受到人们的重视。支持向量机是一种基于结构风险最小化和统计学习理论的机器学习方法,它能有效地解决小样本、非线性、高维数以及局部极值等模式识别问题。此外,研究表明交通流还具有混沌特性。因此,将支持向量机和混沌理论结合应用于短时交通流预测中具有重要意义。本文首先总结了国内外短时交通流预测现状;然后分析证明了交通流数据的混沌特性;最后以此为基础,提出了一种基于支持向量回归机(Support Vector Regression,SVR)和多变量相空间重构的短时交通流预测模型。本文的创新点主要体现在模型的设计原理上,,即本文采用多变量时间序列进行建模。本文主要研究工作如下: ①在介绍PeMS12.3数据库的基础上,分析了交通流基本特征参数(交通流量、占有率和平均速度),研究了交通流数据的预处理方法,并完成了对实测交通流数据的预处理:缺失或错误数据的预处理、降噪处理。 ②在概述混沌理论的基础上,介绍了多变量相空间重构理论,并完成了对预处理后的交通流数据的实验,得出了交通流量、占有率、平均速度时间序列的嵌入维数和延迟时间,以及实现了多变量相空间重构。 ③在分析交通流混沌特性及其混沌特性判别方法的基础上,对交通流量、占有率、平均速度时间序列进行最大Lyapunov指数的计算,结果验证了这三种序列都具有混沌特性。 ④结合混沌理论及支持向量回归机原理,利用遗传算法对支持向量回归机参数进行优化选取,构建了基于多变量相空间重构的SVR短时交通流预测模型,提出了交通流预测流程,给出了预测评价指标(平均绝对误差、平均相对误差、均方误差),最后利用该模型对实测交通流数据进行了实验,同时与基于单变量相空间重构的SVR预测模型进行了比较。 实验结果表明:本文提出的基于多变量相空间重构的SVR短时交通流预测模型的平均绝对误差、平均相对误差和均方误差均小于基于单变量相空间重构的SVR预测模型。说明了本文提出的模型预测效果更好,较充分地验证了本文提出的模型能有效地进行短时交通流预测。
[Abstract]:The real-time and accurate prediction of traffic flow is the precondition and key of the application of traffic signal control system and traffic flow guidance system. Its prediction accuracy is directly related to the operation effect of traffic control and traffic guidance. Because of its randomness, time-varying, strong nonlinear and so on. Support vector machine (SVM) is a machine learning method based on structural risk minimization and statistical learning theory. The problem of pattern recognition such as high dimension and local extremum. In addition, the study shows that the traffic flow also has chaotic characteristics. It is of great significance to apply support vector machine and chaos theory to short-term traffic flow prediction. Firstly, this paper summarizes the current situation of short-term traffic flow prediction at home and abroad. Then the chaotic characteristics of traffic flow data are analyzed and proved. Finally, a support Vector Regression based on support vector regression is proposed. SVR) and multi-variable phase space reconstruction of short-term traffic flow prediction model. The innovation of this paper is mainly reflected in the design principle of the model. That is, this paper uses multivariable time series to model. The main research work of this paper is as follows: 1 based on the introduction of PeMS12.3 database, the basic characteristic parameters of traffic flow (traffic flow, occupancy rate and average speed) are analyzed, and the preprocessing method of traffic flow data is studied. The preprocessing of the measured traffic flow data is completed: the missing or wrong data preprocessing and the noise reduction. 2 on the basis of summarizing chaos theory, the theory of multi-variable phase space reconstruction is introduced, and the experiment of pre-processing traffic flow data is completed, and the traffic flow and occupation rate are obtained. The embedding dimension and delay time of average velocity time series and the reconstruction of multivariable phase space are realized. 3 on the basis of analyzing the chaos characteristic of traffic flow and its distinguishing method, the maximum Lyapunov exponent is calculated for traffic flow, occupation rate and average velocity time series. The results show that the three sequences are chaotic. 4 combined with chaos theory and support vector regression machine principle, the parameters of support vector regression machine are optimized by genetic algorithm, and the SVR short-term traffic flow prediction model based on multi-variable phase space reconstruction is constructed. The flow chart of traffic flow prediction is put forward, and the evaluation indexes (mean absolute error, mean relative error, mean square error) are given. Finally, the model is used to test the measured traffic flow data. At the same time, it is compared with the SVR prediction model based on single variable phase space reconstruction. The experimental results show that the proposed SVR short-time traffic flow prediction model based on multi-variable phase space reconstruction has an average absolute error. The average relative error and mean square error are smaller than the SVR prediction model based on single-variable phase space reconstruction. It is fully verified that the proposed model can effectively predict short-time traffic flow.
【学位授予单位】:重庆交通大学
【学位级别】:硕士
【学位授予年份】:2014
【分类号】:U491.112
【参考文献】
相关期刊论文 前10条
1 李英,刘豹,马寿峰;交通流时间序列中混沌特性判定的替代数据方法[J];系统工程;2000年06期
2 杨世坚,贺国光;基于模糊C均值聚类和神经网络的短时交通流预测方法[J];系统工程;2004年08期
3 王进;史其信;;短时交通流预测模型综述[J];中国公共安全(学术卷);2005年01期
4 王晓,隽志才,朴基男,贾洪飞;局部比较的变点统计理论及其在交通流突变研究中的应用[J];公路交通科技;2002年06期
5 刘静,关伟;交通流预测方法综述[J];公路交通科技;2004年03期
6 孙占全;潘景山;张赞军;张立东;丁青艳;;基于主成分分析与支持向量机结合的交通流预测[J];公路交通科技;2009年05期
7 倪利华;陈笑蓉;;ARIMA模型结合小波去噪的贵阳城市交通流预测[J];贵州大学学报(自然科学版);2011年05期
8 林鑫;王晓晔;王卓;张德干;;基于蚁群聚类算法的RBF神经网络交通流预测[J];河北工业大学学报;2010年03期
9 曹成涛;徐建闽;;基于PSO-SVM的短期交通流预测方法[J];计算机工程与应用;2007年15期
10 陆琳;张虹;;城市短时交通流预测仿真研究[J];计算机仿真;2012年05期
本文编号:1463842
本文链接:https://www.wllwen.com/kejilunwen/jiaotonggongchenglunwen/1463842.html