基于聚类分析和Att-Bi-LSTM的舰船航迹预测方法

The Method of Ship Track Prediction Based on Cluster Analysis and Att-Bi-LSTM

针对应用传统时间序列预测方法对舰船航迹进行预测时,航迹特征提取不充分且预测精确度和预测稳定性不够理想的问题,采用“先聚类分析,后预测”的思想,并综合考虑航迹多维特征,提出一种新的舰船航迹预测方法,先将历史航迹通过压缩、匹配和聚类等三个阶段聚成不同类别,后在双向长短期记忆网络中加入注意力机制,分别在不同类别航迹中构建舰船航迹预测模型,对航迹的航速、经度、纬度和航向等属性进行预测。在包含民用和军用航迹的数据集上设计对比实验,结果表明所提方法在预测精确度和预测稳定性两个方面较现有方法均得到提高。

In view of the problem that the extraction of track features is not adequate,and the precision and stability are not ideal while using the traditional time series prediction method to predict ship track,a new method of ship track prediction is proposed with the thought of cluster analysis before prediction and multi-dimensional features.Firstly,the historical track was clustered into different categories through three stages:compressing,matching,and clustering.Then,the attention mechanism was put into a bidirectional long short-term memory network.Finally,ship track prediction models were established in different categories to predict the ship's speed,longitude,latitude,and course.The comparison experiment was designed on the data set containing the civil and military track data and the results show that the precision and stability are improved compared with existing methods.