基于改进GMDH算法的路口短时交通流量预测  

Short-term traffic flow forecasting on grossroads based on improved group method of data handing

作　　者：王明月[1] 王晶[1] 齐瑞云[1] 陈复扬[1] 

机构地区：[1]南京航空航天大学自动化学院,南京210016

出　　处：《计算机应用》2015年第A01期101-103,134,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61374116);中央高校基本科研专项基金资助项目(2014202)

摘　　要：城市交通是一个复杂的大系统,实时而准确的短时交通流量预测,可以为城市交通诱导和控制提供科学支持。针对GMDH算法建模泛化能力差的问题,结合集成学习的思想对GMDH算法进行改进,并将改进的算法应用到短时交通流量模型的构建中。结果表明,该方法可以有效地对短时交通流量进行预测,建模平均相对误差为1.10%,预测相对误差为0.58%。The urban traffic is a complex large system, actual and accurate traffic flow prediction can provide scientific support for urban traffic guidance and control. Ensemble learning is introduced to improve the general ability of classical Group Method Of Data Handing （ GMDH ） algorithm. The short-term traffic flow model was built based on improved GMDH algorithm. Experimental results indicate that the average relative error of the model is 1. 10%, and the relative error of prediction is 0. 58%. Thus, this model is an efficient method to the short-term traffic flow forecasting.

关 键 词：智能交通系统 短时 交通流量 GMDH 预测 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]