基于SVR对交通流中线性关联关系的分析与研究  

Linear correlative analysis and research of traffic flow based on SVR

作　　者：曹来成[1] 梁浩[1] 韩薇[1] 董胜[1] 

机构地区：[1]兰州理工大学计算机与通信学院,兰州730050

出　　处：《计算机应用研究》2015年第2期419-422,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(60972078);甘肃省自然科学基金资助项目(0916RJZA015)

摘　　要：针对断面交通检测数据往往存在着错误、缺失、包含较多噪声等问题,提出了一种基于支持向量回归机的数据预处理方法。先将相邻路段的数据运用线性回归思想筛选、重组,添加到支持向量回归机的数据集中,然后对相邻路段与预测路段之间线性关系进行实时的、动态的分析和计算,从而避免了数据丢失,既有效地压缩了训练集特征数,提高了计算效率,也提高了模型的泛化能力。实验结果表明,对比未作预处理的SVR模型,改进后的模型拟合度提高了25倍,均方误差也明显减小。To solve the traffic data loss and low computational efficiency of many models in the field of prediction of traffic flow, this paper proposed a data preprocessing method based on support vector regression machine. Firstly, it filtered and recombined the data of adjacent sections based on the idea of linear regression prediction, and then the processed data was added to the data set of support vector regression machine. Secondly, it analyzed and calculated the linear relationship between adjacent sections and forecast section in real time. Thus avoided the loss of data, and compressed characteristics of training sets effectively,so that improved the computational efficiency and the generalization ability of the model. The experimental results show that the improved model enhances the R-Squared by 25 times comparing with the none-pretreatment model, and mean square error reduces obviously.

关 键 词：交通流预测 支持向量回归机 数据预处理 相邻路段 线性关系 

分 类 号：TP391.77[自动化与计算机技术—计算机应用技术]