基于鲸鱼算法优化Attention-Bi LSTM模型的短期船舶流量预测

Short Term Ship Flow Prediction Based on Whale Algorithm Optimized Attention-BiLSTM Model

提出了一种基于双向长短期记忆网络的短期船舶流量预测方法,以历史船舶流量数据为输入,提取反映船舶交通流复杂动态变化的高维特征。引入注意机制,通过映射权重和学习参数矩阵,对网络的隐含状态赋予不同的权重,减少历史信息的损失,增强重要信息的影响。使用鲸鱼算法对模型进行优化,以最小化Attention-Bi LSTM网络期望输出与实际输出之间的均方差为目标,寻找网络超参数,使得网络的误差最小。对长江江苏段苏通大桥断面船舶交通流量进行预测试验,模型拟合优度达到98.8%,并与其他神经网络模型进行对比,试验结果表明文中提出的模型具有更高的预测精度。

A short-term ship flow prediction method based on two-way long-term and short-term memory network is proposed.Taking the historical ship flow data as the input,the high-dimensional features reflecting the complex dynamic changes of ship traffic flow are extracted.The attention mechanism is introduced to give different weights to the implicit state of the network by mapping weights and learning parameter matrix,so as to reduce the loss of historical information and enhance the influence of important information.The whale algorithm is used to optimize the model.Aiming at minimizing the mean square deviation between the expected output and the actual output of the Attention-BiLSTM network,the super parameters of the network are found to minimize the error of the network.The prediction test of ship traffic flow of Sutong bridge section in Jiangsu section of the Yangtze River is carried out,and the goodness of fit of the model is 98.8%.Compared with other neural network models,the test results show that the model proposed in this paper has higher prediction accuracy.