Multivariate time sequence similarity search method based on variable correlation

Abstract

The invention discloses a multivariate time sequence similarity search method based on variable correlation. The method comprises the following steps of firstly, normalizing a pre-acquired multivariate time sequence; secondly, calculating a Pearson correlation coefficient among query sequence variables; on the basis of Pearson correlation, clustering by using OPTICS, and clustering variables with similar trends into one class; then, extracting a central sequence of each class after clustering, and representing the class by using the central sequence; and finally carrying out PAA feature extraction on all the center sequences, and obtaining a feature representation sequence of the multivariate time sequence; and calculating the DTW distance between the multivariate time sequences after feature representation, and finding out similar sequences. According to the method, the feature sequences which are concise and reserve original data features are obtained through variable clustering, and the distance between the feature sequences is calculated by using the multi-dimensional DTW, so that the retrieval efficiency and the query accuracy are improved.

Description

technical field [0001] The invention belongs to the field of information technology, in particular to a multivariate time series similarity search method based on variable correlation. Background technique [0002] Time series is a collection of observations arranged in chronological order, which widely exist in various fields of the real world, including finance, meteorology, medical treatment, engineering, etc. With the continuous development of science and technology, time series data is becoming more and more abundant, and time series similarity search is of great significance to time series prediction, classification, clustering, knowledge discovery, etc. Therefore, time series similarity search is more and more researched readers' attention. [0003] At present, there are many studies on univariate time series similarity measurement, and the research results are relatively rich, but there are not many studies on multivariate time series similarity measurement. The mai...

AI Technical Summary

Problems solved by technology

[0005] However, most of the existing multivariate time series similarity search methods cannot balance efficiency and accuracy. At the same time, some dimensionality reduction methods still lack interpretability. It is necessary to have a multivariate time series similarity search method

Images