基于Attention的AIS数据段相似性搜索模型  

Attention-based Model for Similarity Search Using AIS Data Segments

作　　者：李丽 赖俊星 LI Li;LAI Jun-xing(91001 Army of PLA, Beijing 100072;China Institute of Marine Technology & Economy, Beijing 100072, China)

机构地区：[1]中国人民解放军91001部队通频室,北京100072 [2]中国船舶工业综合技术经济研究院,北京100072

出　　处：《指挥控制与仿真》2022年第1期86-91,共6页Command Control & Simulation

摘　　要：自动识别系统(Automatic Identification System,AIS)数据中蕴含着与船舶交通相关的时间序列信息。基于AIS数据段的相似性搜索也成为智能船运领域的研究热点。AIS数据属于多维时间序列数据,传统的多维数据序列相似性搜索方法通过将原始数据进行降维,并在降维后的数据空间进行时序数据的相似性搜索。然而,此种方法计算复杂度较高,会极大程度地降低相似性搜索的效率。深度学习模型产生的二进制编码改善了时间序列数据搜索性能。基于注意力(Attention)机制的AIS数据段相似性搜索模型对时间序列数据中的时序信息与参数变量间的潜在关系进行同时获取与学习。基于Attention机制的Seq2Seq的相似性搜索模型结合了时间序列数据与Seq2Seq模型的特点,使用Seq2Seq模型中GRU编码器产生的二进制编码进行AIS时间序列数据的相似性搜索。相应的实验结果也表明这种基于Attention机制的模型具有优异的AIS数据相似性搜索性能。Automatic Identification System(AIS)data contains time series information related to ship traffic.Similarity search based on AIS data segment has also become a research hotspot in the field of intelligent shipping.AIS data belong to multivariate time series data.The traditional multidimensional data series similarity search methods reduce the dimension of the raw data and perform the similarity search of the sequential data.However,the computational complexity of this method is high,which will greatly reduce the efficiency of similarity search.The binary encoding generated by deep learning model improves the performance of time series data search.Attention-based model for time series similarity search using AIS Data segments can simultaneously acquire and learn the potential relationship between temporal information and parameter variables in time series data.Seq2Seq similarity search model based on attention mechanism combines the characteristics of time series data and Seq2Seq model,using the binary encoding generated by GRU encoder in Seq2Seq model to search for similarity on AIS data.The corresponding experimental results also show that this model based on the Attention mechanism has excellent similarity search performance on AIS data.

关 键 词：自动识别系统数据 相似性搜索 Seq2Seq 注意力机制 时序信息 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]