基于ARIMA-BIGRU的船舶航迹预测

The Track prediction Method Based on ARIMA-BIGRU Neural Network

为了提高船舶航迹预测的精准度和稳定性,提出一种基于差分自回归移动平均模型和双向循环神经网络的组合预测方法,利用ARIMA模型对航迹进行初步预测,再通过注意力机制优化BIGRU神经网络,对残差序列进行矫正,以船舶经度、纬度、航速、航向、船舶类型作为输入,经度、纬度作为输出,构建船舶航迹预测模型。实验结果表明,该种预测方法的均方误差、平均绝对误差、均方根误差均较小,能够比较准确地预测出船舶航迹。

In order to improve the accuracy and stability of ship track prediction,a combined prediction method of BIGRU neural network based on ARIMA and attention mechanism optimization was proposed.In this method,ARIMA model was used to preliminarily predict the track,and the attention mechanism was used to optimize BIGRU neural network to correct the residual sequence.The ship’s longitude,latitude,speed,course and ship type were used as inputs,and longitude and latitude were used as outputs.The experimental results showed that the mean square error,mean absolute error and root mean square error of this prediction method are relative small,the proposed method can predict the ship track accurately.