摘要
为提高船舶交通流量预测精度,提出一种季节性自回归移动平均(Seasonal Autoregressive Integrated Moving Average,SARIMA)模型和BP神经网络的误差校正集成模型。以深圳港2011—2017年的数据为研究样本,对原始数据进行预处理,构建最优SARIMA模型,以该模型求出的残差序列作为BP神经网络的输入,将两个模型预测结果进行整合,得到集成模型的预测结果。试验结果表明:该误差校正集成模型与两个单一模型相比,体现出船舶交通流量数据的季节性特征,具有较好的预测精度,为港口船舶交通流量预测提供一种更为有效的方法。
The seasonal autoregressive integrated moving average model is integrated with a BP neural network for error correction.The data of Shenzhen Port from 2011 to 2017 are preprocessed and an optimal SARIMA(Seasonal Autoregressive Integrated Moving Average)model is built based on the data.The residual sequence of the output of the SARIMA model is fed to the BP neural network.The combination of the results makes the final prediction.Experiments show that the integrated model reflects the seasonal characteristics of the ship traffic volume well and has better prediction accuracy compared with ordinary design of single model.
作者
赵龙文
苌道方
朱宗良
高银萍
ZHAO Longwen;CHANG Daofang;ZHU Zongliang;GAO Yinping(Institute of Logistics Science & Engineering, Shanghai Maritime University, Shanghai 201306, China)
出处
《中国航海》
CSCD
北大核心
2020年第1期50-55,94,共7页
Navigation of China
基金
国家自然科学基金(71602114)
上海市科委科研项目(16040501500,17595810300)。
