Open AccessTrajectory time series classification algorithm based on conv-olutional self-attention mechanism
Author(s) -
Huamin Yang,
Xiaohui Li,
Zhonglin Liu,
Licai Wang,
Qi Luo
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1961/1/012037
Subject(s) - trajectory , computer science , series (stratigraphy) , mechanism (biology) , feature (linguistics) , algorithm , time series , artificial intelligence , data mining , machine learning , paleontology , philosophy , linguistics , physics , epistemology , astronomy , biology
As a kind of time series data in the field of transportation, trajectory has complex multi-dimensional information. How to fully excavate the multi-dimensional information of the trajectory and analyse its complex dynamic pattern is an important problem to be solved in the classification of trajectory time series. In response to the above problems, this paper improves the existing self-attention mechanism, and proposes a trajectory time series classification algorithm based on convolutional self-attention mechanism. According to the characteristics of the trajectory data, the convolutional self-attention mechanism first extracts the characteristics of the trajectory fragments in the trajectory to obtain the flight status characteristics, and then performs the self-attention operation on these characteristics, so as to achieve a better feature extraction effect. The experimental results show that the trajectory time series classification algorithm based on the convolutional self-attention mechanism has better feature extraction ability and classification accuracy than other existing algorithms.