Accurate forecasting of Earth orientation parameters (EOP) is important for improving the GPS location accuracy and navigation of Earth satellites. EOP time series include periodic components of complex structure. Singular Spectrum Analysis (SSA) is a nonparametric method that is capable of decomposing and forecasting time series with sine-wave components. In the paper, a unified approach to choosing parameters of the SSA forecasting algorithm for EOP time series prediction is proposed. EOP time series data published by IERS in Bulletin 14 C04 are used for 365-days prediction. The forecasts performed by the proposed techniques are compared with predictions taken from available public sources.