Time series classification method and device

Abstract

The invention discloses a time series classification method and device. The method comprises the steps that after a plurality of code words are generated, the first code word and the second code word are determined from the code words, and reconstitution is performed on time series to be tested and sample time series through the first code word and the second code word, wherein the Mahalanobis distance between sub-series of the time series to be tested and the first code word is the shortest; the Mahalanobis distance between sub-series of the sample time series and the second code word is the shortest. Furthermore, by utilizing the Mahalanobis distances between the reconstituted time series to be tested and all the reconstituted sample time series, the classification of the time series to be tested is determined. Compared with the prior art, the problem that Euclidean distances are easily affected by mode characteristic dimensions when serving as similarity measurements is solved. The Mahalanobis distances are introduced to serves as similarity measurements, in this way, the influence on algorithm accuracy by code word relevancy is eliminated while the influence of dimensions is eliminated, and classification accuracy is improved.

Description

technical field [0001] The present invention relates to the field of actual sequence data mining, and more specifically relates to a time series classification method and device. Background technique [0002] A time series is an ordered sequence of various data of a certain phenomenon or statistical index at different time points, arranged in chronological order. [0003] With the advent of the data age, it is especially important to quickly and efficiently classify messy time series. The classification of time series has always been one of the key research directions in the field of time series data mining. The classification of time series can be decomposed into two sub-problems, that is, how to represent the time series (or how to use a certain storage space to accommodate more time series), and how to compare the time series to be tested with the time series of known categories to determine the category of the data to be tested. [0004] At present, segmented vector qu...

AI Technical Summary

Problems solved by technology

However, the Euclidean distance is susceptible to the dimensionality of the pattern features, which may lead to a reduction in classification accuracy

Images