Robust singular spectrum analysis: an empirical study

This paper is aimed at studying robust methodologies for decomposition, filtering and reconstruction of time series, which is relevant for treating events that can cause major effects on the data and on its dependency... [ view full abstract ]

This paper is aimed at studying robust methodologies for decomposition, filtering and reconstruction of time series, which is relevant for treating events that can cause major effects on the data and on its dependency structure. Based on the proposal of Yarmohammadi, M. and Kalantari, M., the interest in this study is to verify the performance of the robust methodology in scenarios which the data are contaminated by additive outliers. Moreover, a new alternative to the same context is suggested by using an M-estimator with the Huber function while applying SSA in financial time series.