Data Analysis: Generalizations of the Local Approximation Method by Singular Spectrum Analysis

Abstract

We study forecasting capabilities of the methods of Singular Spectrum Analysis (SSA) and Local Approximation (LA). A practical implementation of these methods to several time series is described. Details of the algorithms of these methods are discussed. Advantages and disadvantages of SSA and LA are given. On the basis of our results we generalize LA and SSA and propose a new way for time series forecasting including strongly noisy signals. This allowed us to extend the range of application of LA in comparison with the known standard scheme. For the problem of forecasting, the accuracy enhancement of the numerical computations is discussed.