基于群智能算法优化SVR的短时交通流预测

发布时间：2018-02-04 17:57

本文关键词： 短时交通流预测支持向量回归人工鱼群算法参数选择混沌初始化　出处：《大连理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：智能交通系统是缓解道路交通拥堵、减少交通事故和提高交通运行效率的重要应用系统。实时准确可靠的交通流量预测是实现智能交通系统控制和诱导的关键内容,具有重大的理论研究和实际应用价值。本文以短时交通流量预测为研究主题,总结了短时交通流预测的研究现状,在学习交通流预测原理和支持向量回归(Support Vector Regression, SVR)理论的基础上,对基于SVR的短时交通流预测模型中参数选择问题进行了探讨和研究,运用群智能优化方法进行最优参数选择,并且仿真实际数据来验证提出的预测模型。本文的主要工作如下：1.对人工鱼群算法优化支持向量回归的参数选择模型进行研究。针对支持向量回归的惩罚系数、不敏感损失系数和核函数参数的选择对回归算法的预测精度的重要影响,结合交通流数据特征,本文运用人工鱼群算法对支持向量回归参数进行优化选择,同时引入人工鱼群算法中感知视野和移动步长参数的自适应搜索机制,建立了基于人工鱼群算法优化支持向量回归的短时交通流预测模型。实际数据的仿真实验和模型的对比结果表明了提出的回归预测模型的可行性和有效性。2.对混合粒子群人工鱼群算法优化支持向量回归的参数选择模型进行研究。在人工鱼群算法优化支持向量回归的预测模型的研究基础上,为解决人工鱼群算法中的初始参数较多问题以及步长因子设置对寻优性能的影响,本文提出采用粒子群优化算法对人工鱼群算法进行改进,减少了步长因子对人工鱼群算法影响,并且引入混沌机制初始化人工鱼群位置信息,从而对支持向量回归进行参数选择,建立了基于混合粒子群人工鱼群优化支持向量回归的短时交通流预测模型。通过仿真实验分析,提出的混合优化预测模型比单一的粒子群和人工鱼群算法优化支持向量回归预测模型有更优的预测性能。
[Abstract]:The Intelligent Transportation system (its) is designed to ease traffic congestion on roads. It is an important application system to reduce traffic accidents and improve traffic efficiency. Real time accurate and reliable traffic flow prediction is the key to realize intelligent traffic system control and guidance. It has great theoretical research and practical application value. In this paper, short-term traffic flow forecasting as the research topic, summarized the research status of short-term traffic flow forecasting. On the basis of studying the theory of traffic flow prediction and support vector regression support Vector Regeneration (SVR). The problem of parameter selection in short-term traffic flow forecasting model based on SVR is discussed and studied, and the optimal parameter selection is carried out by using swarm intelligence optimization method. And simulate the actual data to verify the proposed prediction model. The main work of this paper is as follows:. 1. The parameter selection model of support vector regression optimization based on artificial fish swarm algorithm is studied. The penalty coefficient of support vector regression is studied. The selection of insensitive loss coefficient and kernel function parameters has an important influence on the prediction accuracy of regression algorithm. Combined with the characteristics of traffic flow data, the artificial fish swarm algorithm is used to optimize the parameters of support vector regression. At the same time, the adaptive search mechanism of perceptual visual field and moving step size parameters is introduced in artificial fish swarm algorithm. A short-term traffic flow forecasting model based on artificial fish swarm optimization support vector regression is established. The simulation results of the actual data and the comparison of the models show that the proposed regression forecasting model is feasible and effective. The parameter selection model of support vector regression optimization based on hybrid particle swarm optimization algorithm is studied, and the prediction model of support vector regression optimization based on artificial fish swarm algorithm is studied. In order to solve the problem of more initial parameters in artificial fish swarm algorithm and the effect of step size factor setting on the optimization performance, the particle swarm optimization algorithm is proposed to improve the artificial fish swarm algorithm. The influence of step size factor on artificial fish swarm algorithm is reduced, and chaotic mechanism is introduced to initialize the position information of artificial fish swarm, so as to select the parameters of support vector regression. A short-term traffic flow prediction model based on hybrid particle swarm artificial fish swarm optimization support vector regression was established and analyzed by simulation experiments. The proposed hybrid optimization prediction model has better prediction performance than the single particle swarm optimization and artificial fish swarm optimization support vector regression prediction model.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;TP18

【参考文献】