Grouping singular spectrum analysis components via mixed integer quadratic programming

Abstract

This paper proposes a method for grouping singular spectrum analysis (SSA) components via the empirical mode decomposition (EMD) approach. To perform the grouping, the total number of the groups of the singular spectrum analysis components is equal to the total number of the intrinsic mode functions (IMFs). The SSA components are assigned to the group where the 2-norm between the IMFs and the grouped SSA components is minimum. It can be formulated as a mixed integer quadratic programming (MIQP) optimization problem. After applying our proposed grouping method to obtain grouped SSA components, the grouped components enjoy the advantages of both the singular spectrum analysis approach and the empirical mode decomposition approach. The proposed technique was tested on electrocardiogram signals showing that the useful information can be concentrated in a few groups, thus most of useful information can be extracted easily.