Analysis And Application Of Multivariate Time Series Similarity

The analysis of time series has important research and application value.This paper studies and analyzes the multivariate time series similarity measurement method and multivariate time series anomaly detection,mainly involving the research of the similarity measurement method and anomaly detection algorithm.In the analysis of multivariate time series similarity,a standard data set simulation experiment is selected to compare and analyze the performance of several common multivariate time series similarity measurement methods.In the study of application of multivariate time series similarity,the similarity measure for traditional outlier detection algorithm ignores the factors related to time bending,unable to ensure the accuracy of multivariate time series in anomaly detection results.To solve this problem,this paper optimizes and improves the traditional outlier detection algorithm,and proposes a improve distance-based outlier detection algorithm.And selecting multivariate time series standard data to verify the effectiveness and superiority of the improved distance-based outlier detection algorithm.Regarding the influence of the weight of the time series data set on the detection results,paper further proposes an outlier detection algorithm based on weighted optimal distance.Firstly,the dimension of each multivariate time series is reduced by principal component analysis.Secondly,the distance between the eigenmatrix of each multivariate time series and its orthogonal coordinate system is calculated by the included Angle formula,and the weighted minimum distance of the multivariate time series is obtained by using the Hungarian algorithm.Finally,an outlier algorithm based on weighted optimization distance is constructed to achieve the anomaly detection of multivariate time series.Four kinds of multivariate time series data sets and transformer measured vibration data sets are selected to complete the experimental simulation of the outlier detection algorithm based on weighted optimized distance.At the same time,the algorithm is compared with traditional outlier detection algorithms and Riemannian manifold algorithms.It shows that the algorithm proposed in this paper has better detection performance.