IDENTIFICATION OF THE STRUCTURE OF NONSTATIONARY TIME SERIES WITH THE SINGULAR SPECTRUM ANALYSIS METHOD

A.A. Chistyakova, B.V. Shamsha

This paper represents the method of Singular Spectrum Analysis (SSA) to identify the structure of non-stationary time series. The purpose of this method is the selection of the time series components, such as trend and periodic component. Solution of this problem is necessary for constructing the model of time series and determination of the masked dependences. The analysis of the structure of nonstationary time series of prices of the sugar is carried out Recommendations on the choice of parameters of SSA to identify the components of time series, which cannot be reduced to a uniform are given. The model of nonstationary time series taking into account the components of trend and periodicals is built.

Key words: nonstationary time series, identification of the model, singular spectrum analysis, principal components.