中文核心期刊
CSCD来源期刊
中国科技核心期刊
RCCSE中国核心学术期刊

重庆交通大学学报(自然科学版) ›› 2019, Vol. 38 ›› Issue (08): 7-12.DOI: 10.3969/j.issn.1674-0696.2019.08.02

• 交通+大数据人工智能 • 上一篇    下一篇

基于极限学习机的船舶航行行为预测

谢新连,陈紫薇,魏照坤,赵瑞嘉   

  1. (大连海事大学 物流研究院,辽宁 大连 116026)
  • 收稿日期:2018-04-24 修回日期:2018-04-20 出版日期:2019-08-01 发布日期:2019-08-01
  • 作者简介:谢新连(1956—),男,辽宁大连人,教授,博士,主要从事交通运输规划与管理方面的研究。E-mail:xxlian@dlmu.edu.cn。 通信作者:陈紫薇 (1994—),女,辽宁沈阳人,硕士研究生,主要从事交通运输规划与管理方面的研究。E-mail:candice0527@sina.com。
  • 基金资助:
    国家重点研发计划项目 (2017YFC0805309);国家自然科学基金项目(61473053);中央高校基本科研业务费专项资金资助项目 (3132019303)

Ship Navigation Behavior Prediction Based on Extreme Learning Machine

XIE Xinlian, CHEN Ziwei, WEI Zhaokun, ZHAO Ruijia   

  1. (Logistics Research Institute, Dalian Maritime University, Dalian 116026, Liaoning, P. R. China)
  • Received:2018-04-24 Revised:2018-04-20 Online:2019-08-01 Published:2019-08-01

摘要: 为提高海上监控系统效率,有效预测船舶航行行为,建立了基于极限学习机的船舶航行行为预测模型。该模型针对航行状态的改变(主要为转向或变速),采取自动调整采样周期的方法更精准的训练网络,从而提高对船舶行为的预测精度。最后,利用琼州海峡的船舶自动识别系统(Automatic Identification System, AIS)信息将设计的预测模型与现有的灰色关联和BP模型进行对比。仿真结果表明:设计的算法有效地降低了船舶在转向及变速前后的预测误差;通过曼-惠特尼U检验证明,设计的基于极限学习机的船舶航行行为预测模型相比于传统BP神经网络及灰色关联模型,在预测精度方面具有更大的优势。

关键词: 交通运输工程, 船舶航行行为预测, 极限学习机, AIS信息, 曼-惠特尼U检验

Abstract: In order to improve the efficiency of marine monitoring system and predict ship navigation effectively, a prediction model of ship navigation behavior was established based on extreme learning machine. According to the change of navigation state (mainly for steering or speed change), the proposed model automatically adjusted the sampling period to train the network more accurately, so as to improve the prediction accuracy of ship behavior. Finally, using the information of the Automatic Identification System (AIS) in Qiongzhou Strait, the designed prediction model was compared with the current grey correlation and BP models. The simulation results show that the designed algorithm effectively reduces the prediction errors of ships before and after steering and speed change; through the Mann Whitney U test, it is proved that the proposed ship navigation behavior prediction model based on extreme learning machine has greater advantage in prediction accuracy, compared with the traditional BP neural network and grey correlation model.

Key words: traffic and transportation engineering, ship navigation behavior prediction, extreme learning machine, AIS information, the Mann-Whitney U test

中图分类号: