Research of data mining method on multivariate time series
Wei Wang · Jisuanji gongcheng yu sheji · 2006
The multivariate time series(MTS)dataset is a common data type in various scientific domains.An MTS is usually very high dimensional with its main distinguishing characteristic being the inter-correlations and/or interdependencies among its variables.Consequently,MTS may not be easily broken into multiple univariate time series and is treated as a whole.A time series pattern mining method based on similarity of MTS,finds some similar multivariate time series from history data,and then stored as time series pattern.Use extending SVD-based similarity measures for MTS datasets by representing an MTS as a matrix.Then get the longer time series seg-ment through bottom-to-up merge between adjacent segments to improve the ability of real time forecast.Several experiments performed on the earthquake auspice datasets.