高速公路短时交通流量预测的改进非参数回归算法  

An Improved Nonparametric Regression Algorithm for Short-term Expressway Traffic Flow Forecasting

作　　者：孙棣华[1,2] 李超[1,2] 廖孝勇[1,2] 

机构地区：[1]重庆大学自动化学院,重庆400044 [2]重庆大学信息物理社会可信服务计算教育部重点实验室,重庆400044

出　　处：《公路交通科技》2013年第11期112-118,共7页Journal of Highway and Transportation Research and Development

基　　金：中国工程院重点咨询项目(2012-xz-22);教育部博士点基金项目(20120191110047);重庆市科技攻关项目(cstc2012gg yyjsB30001)

摘　　要：针对非参数回归短时交通流量预测算法的状态向量选取问题,基于高速公路交通流量在空间上演变趋势明显的特点,提出交通流量预测的改进非参数回归算法。引入各上游断面车流到达当前断面的行程时间作为状态向量选取的依据,并根据各上游断面影响程度的不同,调整相似机制的计算方法。利用渝武高速公路微波检测器数据对该模型进行验证。结果表明,改进的非参数回归算法克服了固定状态向量定义不能满足同一断面不同交通状态的缺点,对各种交通状态具有更好的适应性,预测精度更高。Aiming at the state vector selection in nonparametric regression for short-term traffic flow forecasting, an improved nonparametric regression algorithm for traffic flow forecasting is proposed based on the characteristic of obvious spatial evolution trend of expressway traffic flow. Taking the travel time of traffic from upstream sections to current section as the basis of dynamic state vector selection, the calculation method of similar mechanism is adjusted according to the effect of each upstream section. The model is tested and verified with the microwave detectors data of Chongqing-Wusheng expressway. The result indicates that the improved nonparametric regression algorithm has overcome the shortcoming that the fixed state vector cannot meet different traffic conditions at the same section, and has a better adaptability to various traffic states and a higher precision of prediction.

关 键 词：交通工程 短时交通流量预测 非参数回归算法 状态向量 行程时间 相似机制 

分 类 号：U491.112[交通运输工程—交通运输规划与管理]