Time series trend dynamic segmentation method based on central point

Abstract

The invention discloses a time sequence trend dynamic segmentation method based on a central point, and aims to establish an effective index for a time sequence database so as to better classify a time sequence. In the field of pattern classification, normal and abnormal conditions of electrocardiogram center hops are automatically distinguished, trend characteristics of a time sequence need to bequalitatively researched, and a time sequence trend segmentation result is dynamically determined on the basis that the overall trend and measurement errors of data are reserved. The method comprisesthe steps of firstly, obtaining effective central points according to a parameter change rule of a time sequence; according to a generation mode of a central point, obtaining dynamic segments of thetime sequence, and determining starting and ending positions and interval trend characteristics of the segments; and finally, establishing a trend segment index table according to the trend index connection of each segment of time sequence. According to the method, the time sequence index is established on the basis of rapid segmentation, the interval trend in the segment is effectively described,and an index table with global capacity is provided for similarity pattern matching research in pattern classification mining.

Description

technical field [0001] The invention belongs to the field of pattern classification in artificial intelligence, and in particular relates to a pivot point-based dynamic segmentation method for time series trends. Background technique [0002] A pattern classification algorithm was used to distinguish between normal and abnormal heartbeats in the ECG. First, the heartbeat is collected to obtain the electrocardiogram, the time series of each heartbeat is extracted, and the interpolation method is used to ensure that the time series of each heartbeat is of equal length. To analyze the time series and mine the similarity patterns, it is necessary to qualitatively study the trend characteristics of the time series. Data in a time series changes accordingly over time. The computational complexity of mining and modeling directly on the original sequence is high, and the accuracy of the obtained results is not high. The trend feature analysis of time series is an important part o...

AI Technical Summary

Problems solved by technology

[0004] In order to improve the versatility of the algorithm itself and reduce the time complexity, many corresponding algorithms have emerged, basically using key points, feature points and trend turning points for segmented linear representation, etc. These methods are useful for time series trend extraction and data compression. It has a certain effect, but the feature points and their evaluation functions are confirmed by the relationship between adjacent multiple points in the interval. It belongs to the local analysis method and cannot reflect the overall trend. The performance of the algorithm is greatly affected by the input parameters, and the result is unstable.

Images