Efficient time sequence segmentation method

Abstract

The invention provides an efficient time sequence segmentation method. The trend feature is important information of a time sequence, linear representation of the time sequence is achieved by segmenting the time sequence, and therefore extraction of the trend feature of the time sequence is an important direction in the field of time sequence research. The time sequence segmentation method comprises on-line segmentation and off-line segmentation. The off-line segmentation method has the characteristic of high representation precision. The on-line segmentation method has the characteristic of on-line processing, and can still continue the optimization of the time sequence segmentation method. According to the method, an on-line segmentation method FSW algorithm and an off-line segmentation method BU algorithm are integrated, and a FSWAB method is provided, so that the new method can have the advantages of the two algorithms, and efficient on-line segmentation and accurate representation can be carried out on the time sequence. The inventor carries out empirical analysis by using a UCR_Archive data set provided by a riverside branch of California University, and the result shows that the method can efficiently segment the time sequence online and accurately represent the trend characteristics of the time sequence.

Description

technical field [0001] The invention relates to a time series segmentation method, in particular to an online segmentation and linear representation method of the time series. Background technique [0002] A time series is an ordered collection of data that changes over time and is related to each other. With the advent of the era of big data, the amount and complexity of time series data are increasing day by day, and gradually have the characteristics of big data. Online segmentation and piecewise linear representation of time series are of great significance for big data mining research. [0003] Piecewise linear representation (PLR) is an important method in the approximate representation of time series. It has been widely used because of its simplicity and intuitiveness, high processing efficiency, and the representation results conform to human visual experience. The classic piecewise linear representation method According to its processing method, it can be divided ...

AI Technical Summary

Problems solved by technology

In order to solve the problem of low accuracy of the original data representation by the classic SW method, Keogh et al. proposed a method based on sliding window and top-down strategy (sliding window and bottom-up, SWAB), which combines the online segmentation method And the advantages of offline segmentation method, improve the accuracy of fitting, but there is a problem of low processing efficiency

In order to solve the problem of low processing efficiency of the SW method, Liu et al. proposed a method based on the feasible space window (FSW). The FSW method uses the method of slope calculation to update the upper and lower boundaries in real time as the method of judging the segmentation point. The processing speed has been greatly improved, but it has also led to a significant drop in representation accuracy, making it impossible to provide high-quality representation results for time series big data mining research work

Images