基于深度神经网络的船舶交通流预测  

Prediction of Ship Traffic Flow Based on Deep Neural Network

作　　者：索永峰[1] 陈文科 杨神化[1] 陈立媛 SUO Yongfeng;CHEN Wenke;YANG Shenhua;CHEN Liyuan(Navigation College,Jimei University,Xiamen 361021,China)

机构地区：[1]集美大学航海学院,福建厦门361021

出　　处：《集美大学学报（自然科学版）》2020年第6期430-436,共7页Journal of Jimei University：Natural Science

基　　金：国家自然科学基金项目(51579114,51879119);福建省自然科学基金项目(2020J01660,2018J01536)。

摘　　要：为了准确预测船舶交通流量,构建一种利用特定数据集进行船舶交通流量预测的深度学习模型。选定海域横断面,统计规定时间段内穿越该海域横断面的船舶AIS数据,将这些数据筛选后作为数据集。选取GRU(gate recurrent unit)模型最佳结构和参数,对一天内的船舶流量进行预测,并选取LSTM(long short term memory)循环神经网络模型和SAES栈式编码器预测模型作为实验对照组模型,在合理参数范围内对不同参数组合进行实验。实验结果表明,与LSTM模型和SAES(stacked auto-encoders)模型相比,GRU模型预测精度更高,能适应数据规律性较弱的船舶交通流量预测的要求。In order to accurately predict the traffic flow of ships,a depth-learning model for predicting the traffic flow based on a specific data set is constructed:The sea cross section is selected,and the AIS data of ships traversing the sea cross section in a specified period of time are statistically analyzed,which are then filtered and used as a dataset.The LSTM and its improved network GRU are selected as the research object,and the model parameters are adjusted to carry out the comparative analysis of several groups of experiments.The results show that compared with the LSTM model and the SAES model,the GRU model is more accurate.The GRU model is more suitable for the prediction of the ship traffic flow data with weak data regularity.

关 键 词：船舶交通流量 深度学习 LSTM GRU 

分 类 号：U692[交通运输工程—港口、海岸及近海工程]