Robust singular spectrum analysis: An application to the energy sector

Singular spectrum analysis (SSA) is a non-parametric method for time series analysis and forecasting that incorporates elements of classical time series analysis, multivariate statistics, multivariate geometry, dynamical systems and signal processing, and can be seen as an alternative to the classical methods. Although SSA has proved its usefulness and advantages over classical methods, one of the steps of the SSA algorithm is the singular value decomposition (SVD) of the trajectory matrix which, being a least squares method, is highly sensitive to data contamination and the presence of even a single outlier, if extreme, may draw the leading principal component towards itself resulting in possible misinterpretations and in turn lead to bad practical decisions. In this paper we propose a robust alternative to SSA, which uses a robust SVD algorithm, that that overcomes the potential fragility of its classical version when the data are contaminated. The SSA and the robust SSA are compared in terms of quality of the model fit and forecasting. This is done by using Monte Carlo simulations and real data from the energy sector.