Similarity Search Method of Hydrological Time Series based on Fragment Alignment Distance and Dynamic Time Warping

Abstract

In hydrological time series mining, hydrological time series similarity mining is an important aspect. To search hydrological time series more effectively, this paper proposes a fast search method for hydrological time series from the perspective of time series trend characteristics. Based on wavelet transform, feature point extraction, and semantic symbolization of time series, the preliminary candidate set is screened by semantic similarity matching, the first M approximate matching sequences with small fragment alignment distance in the preliminary candidate set are selected, and the first M approximate subsequences are accurately matched by dynamic time warping distance to obtain the final similar sequence. The water level data of the Tunxi station in the Tunxi basin are used in the experiment, and the results show that the proposed method can greatly improve the search efficiency while ensuring accuracy.