Abstract
Singular Spectrum Analysis (SSA) as a nonparametric tool for decomposition of an observed time series into sum of interpretable components such as trend, oscillations and noise is considered. The separability of these times series by SSA means the possibility of such decomposition. Two extensions of SSA, which weaken the separability conditions, are proposed. One of the proposed approaches considers inner products corresponding to oblique coordinate systems instead of the conventional Euclidean inner product. The other method changes the balance of the components by involving the series derivative to avoid component mixing (if any). Each suggested method is supplemented by the algorithm and is demonstrated by real-life examples.

This publication has 2 references indexed in Scilit: