A new approach to identifying the local structure of multidimensional chaotic time series

Abstract

The paper is devoted to the solution of the problems of mathematical supply of decision making during multichannel monitoring of large-scaled systems. The work also deals with space-time dynamics of multidimensional time series of different origins. Highly dynamical chaotic processes whose fine structure cannot be revealed by standard spectral methods are regarded. Technologies for dimension reduction based on data matrix representation on the first singular basis and multiple regression in projections’ space are developed.