Time sequence variable-length motif mining method based on suffix tree

Abstract

The invention relates to a time sequence variable-length motif mining method based on a suffix tree. The method comprises the following steps: performing mode representation based on a slope, setting a change rate threshold, and extracting all edge points to obtain an edge point set; using the edge points of the edge point set to construct a suffix tree, and using the suffix tree to statistically determine frequency of edge point subsequences, wherein the edge point subsequence with the maximum frequency is a frequent pattern; mapping the frequent pattern back to an original time sequence, and recording position of variable-length motifs; and according to the positions of the variable-length motifs, calculating Matrix Profile values between the variable-length motifs, wherein the motif with the minimum Matrix Profile value is an effective motif. According to the invention, extraction of an effective motif is added, so that the problem of low motif discovery precision caused by symbolization hiding extreme point information is solved, and the time sequence variable-length motif mining precision is improved.

Description

technical field [0001] The present application relates to the technical field of information processing, in particular to a time series variable length motif mining method based on a suffix tree. Background technique [0002] Time series data mining belongs to the category of data mining. Its main goal is to discover meaningful information from time series data, and it needs to complete tasks such as clustering, classification, similarity search, anomaly detection, and motif mining. Among them, time series motif mining is to find recurring unknown patterns in time series without any prior information about their location or shape. In addition, time series motif mining is not only applicable to one-dimensional or multi-dimensional data, but also applicable to different types of sequence data, such as spatial sequence data, time series data, and stream data. And time series motif mining technology has also been applied in many fields such as genetics, medicine, mathematics, m...

AI Technical Summary

Problems solved by technology

For example, the motif discovery by random projection is to calculate the mean value of the sequence, and to find frequent patterns after symbolic representation according to the mean value, so it can only ensure that the overall change trend of the motifs is the same, and cannot ensure the similarity between the results, resulting in longer time series motifs Mining accuracy is relatively low

Images