基于广义回归神经网络的船舶交通量预测模型  

A Marine Traffic Flow Forecasting Model Based on Generalized Regression Neural Network

作　　者：刘敬贤[1] 刘振东[1] 周锋[1] 

机构地区：[1]武汉理工大学航运学院,湖北武汉430063

出　　处：《中国航海》2011年第2期74-77,85,共5页Navigation of China

摘　　要：船舶交通量受多种环境与社会因素的影响,使得船舶交通量预测存在复杂性与非线性的特点。在分析现有预测模型和方法不足的基础上,介绍了广义回归神经网络GRNN的基本原理与拓扑结构。不同类型船舶受各类因素影响的程度不同,根据天津港VTS(Vessel Traffic Services)中心提供的船舶交通量数据,按船舶种类将船舶交通量分为六类,利用GRNN神经网络分别进行预测。预测结果表明GRNN神经网络具有很强的非线性拟合能力,有效解决了天津港船舶交通量预测中的小样本问题,提高了整个预测系统的精度与稳定性。Marine traffic flow is influenced by a variety of environmental and social factors,so forecasting of it has complex and non-linear characteristics.Since the impacts of various factors on different kind of ships are different,the ships are divided into 6 categories based on the data of ship traffic flow provided by VTS Center of the Tianjin Harbor.To improve the marine traffic flow forecasting,the basic principle and topology structure of the generalized regression neural network（GRNN） are introduced for forecasting marine traffic flows of different vessel types.Results show that GRNN neural network has good non-linear approximation capability and can solve the small sample problem of statistical data and forecast marine traffic flow of the Tianjin Harbor effectively.The method enhances the prediction precision and stability of forecasting system.

关 键 词：水路运输 船舶交通量 广义回归神经网络 小样本问题 组合预测模型 

分 类 号：U692.3[交通运输工程—港口、海岸及近海工程] TP274[交通运输工程—船舶与海洋工程]