Research On Time Series Representation Method And Anomaly Detection Technology Based On Interval-Sets

With the advent of artificial intelligence and information network era,a large amount of data has been accumulated,and the importance of rational analysis and application of data to human life has become increasingly obvious.As a kind of data form with time attribute,time series data boasts strong representativeness among all kinds of data.However,the study of anomaly detection of time series has always been a hot and challenging research topic due to some major features of time series data such as nonlinearity,high dimension and time attribute.Most scholars mainly focus on the design and optimization of anomaly detection algorithm while studying anomaly detection of time series,and there are few related literatures on time series data representation.On the basis of original data,an effective time series data representation model can reasonably reduce dimensionality,extract data features,improve the anti-noise ability of data,and increase the accuracy and robustness of detection methods.Granular computing,which can represent complex information particles through constructing and processing information,plays an important role in the field of mining and analyzing time series data.The purpose of this paper aims to design a time series representation method based on interval sets and build the corresponding anomaly detection framework through combining existing data representation method of the existing research results and related applications of particle computing in time series representation,and viewing interval information as a form of information granule,realize anomaly detection of time series data.In this paper,a piecewise approximate representation method of time series based on interval sets is proposed.ISPA(interval set piecewise approximation)divides time series into multiple sub-sequences according to the research results of related information particles,and each sub-sequence is granulated into an interval information particle.The time series is divided into several sub-sequences and each sub-sequence is granulated into an interval information granule.The time series data are extracted and expressed by constructing reasonable information granules.Moreover,based on information and particle gradation characteristics and the ISPA representation method,this paper construct a time series representation method based on multi-granularity Interval Sets M-ISPA(Multi-granularity Interval Sets Piecewise Approximation),the method establishes the multi-level expression of time series data by adjusting the information particle size,which makes the original time series information obtained more reasonable description,further improved the stability and the accuracy of time series of anomaly detection.Based on the representation methods of ISPA and M-ISPA,this paper adopts the relevant research of interval operation and interval boundary range optimization,designs an appropriate interval set similarity measurement method,and proposes an anomaly detection algorithm based on this method,which is verified by a large number of experiments.Experimental results show that this method can effectively detect abnormal behavior in time series data,with the accuracy of more than 90%,where greatly reduce the false alarm rate and missed detection rate of abnormal detection.