Segmentation extraction of feature points for time series pattern matching

It is difficult for the common time series pattern matching methods to balance the computational complexity and matching accuracy. To solve this problem, a time series matching method based on segmented extraction of feature points is proposed. Firstly, the feature points on each variable dimension of the time series are extracted and the sequence length is compressed. Then, the quantile matrix is calculated according to the feature sequence, and the similarity of the quantile matrix is measured by Euclidean distance. Finally, the effectiveness of the proposed method is verified on the application data set. Experimental results show that the proposed method can effectively reduce the computational complexity and ensure high matching accuracy. © 2023 Beijing University of Aeronautics and Astronautics (BUAA). All rights reserved.