交通运输系统工程与信息

交通运输系统工程与信息 ›› 2013, Vol. 13 ›› Issue (3): 33-39.

• 智能交通系统与信息技术 • 上一篇    下一篇

基于生长自组织神经网络群的交通流预测

吕进*1，赵祥模１，樊海玮１，旺乃姆２   

  1. 1. 长安大学 信息工程学院，西安 710064; 2. 多伦多大学 电子与计算机工程系， 加拿大 多伦多 M5S3G4
  • 收稿日期:2012-12-18 修回日期:2013-03-14 出版日期:2013-06-25 发布日期:2013-07-02
  • 作者简介:吕进（1972-），男，陕西西安人，副教授，工学博士.
  • 基金资助:

    国家自然科学基金资助项目（50908017）；中国博士后科学基金项目（20090451363）；中央高校基本科研业务费专项资金项目（CHD2010TD001，CHD2011ZD015）.

Traffic Flow Forecasting Based on Growing Self-organized Neural Network Group

LV Jin１, ZHAO Xiang-mo１, FAN Hai-wei１, WONHAM W M２   

  1. 1. School of Information Engineering, Chang’an University, Xi’an 710064, China; 2.Department of Electrical and Computer Engineering, University of Toronto, Toronto, Ontario M5S 3G4, Canada
  • Received:2012-12-18 Revised:2013-03-14 Online:2013-06-25 Published:2013-07-02

PDF

被引次数

4

摘要：

为了提高城市交通流预测神经网络方法的快速动态学习能力，提出了一种生长自组织神经网络群，将复杂的神经网络个体分解为多个训练简单的神经网络群组，并利用设计的动态生长自组织算法来避免神经网络在学习新知识的时候对已有知识造成破坏，同时保持整个群工作的高效稳定，规模不过度扩张.该神经网络群尝试解决神经网络的一次性学习问题，具有动态知识增殖学习能力和更强的错误自修复能力及系统适应灵活性.仿真结果表明，这一方法能够更精确地实现函数逼近和城市交通流自适应动态预测，适用于需要不断快速动态学习的复杂系统.

关键词: 智能交通, 交通信息工程及控制, 交通流预测, 人工神经网络, 生长自组织神经网络

Abstract:

To enhance the capacity of dynamic study and realtime forecasting on urban traffic flow, this paper proposes a type of growing selforganized neural network group (GSNNG). A complex artificial neural network (ANN) is introduced into some easytrained ANNgroups, and the dynamicgrowing selforganized algorithm is adopted to avoid the ANN damages to the acquired knowledge when it learns some new ones. The algorithm is able to maintain the stability of the whole ANNgroups, as well as the efficiency and a reasonable scaleconfined. The GSNNG solves the ANN’s problem that new knowledge affects on the old ones, which had more dynamic knowledgeincreasable, errorsselfrepairing and adapting capacity. Simulation results show that the GSNNG produces higher forecasting precision and stronger dynamic performance in systemidentification and traffic flow forecasting. The method is fit to the complex systems which need continual dynamicstudy.

Key words: intelligent transportation, traffic information engineering and control, traffic flow forecasting, artificial neural network, growing selforganized neural network

中图分类号: