多模态的交通流量预测模型  

Multimode traffic volume prediction model

作　　者：董宏辉[1] 孙晓亮[1] 贾利民[1] 秦勇[1] 

机构地区：[1]北京交通大学轨道交通控制与安全国家重点实验室,北京100044

出　　处：《吉林大学学报（工学版）》2011年第3期645-649,共5页Journal of Jilin University:Engineering and Technology Edition

基　　金：'863'国家高技术研究发展计划项目(2006AA11Z231);北京市科技计划重点项目(D07020601400707)

摘　　要：针对交通状态的多模态性,提出了多模态的交通流量预测方法。引用道路服务水平将交通状态分为6级(类)模态,并研究了不同模态与流量之间的对应关系。多模态的交通流量预测模型根据历史数据判断交通模态的改变情况,在整合自回归移动平均模型(ARIMA)预测的基础上,利用模态修正函数动态调整ARIMA预测中产生的时滞误差。以实际交通流数据为样本进行的实验分析结果表明,多模态的交通流量预测模型在单步、多步长预测中比ARI-MA模型有更高的预测精度。A muhimode traffic volume prediction model was proposed in the light of the muhimode behavior of the traffic state. The traffic state was divided into 6 modes according to the level of service, and the relation between the traffic volume and the mode was studied. The proposed model estimates the variation of the traffic mode based on the historical data. On the basis of the ARIMA prediction model, the time-delay error in the ARIMA prediction was adjusted dynamically using a dynamic correction function. The experiments and analyses using the actual traffic flow data as a sample showed that the muhimode traffic volume prediction model provides a better performance than the ARIMA model in single-step and multi-step predictions.

关 键 词：交通工程 交通流预测 交通状态 多模态模型 

分 类 号：U491[交通运输工程—交通运输规划与管理]