Study of automatic choice of parameters for forecasting in singular spectrum analysis

Abstract

Singular spectrum analysis (SSA) is a popular tool for analysing and forecasting time series. The SSA forecasting algorithms have two parameters which should be chosen by the researcher or using the so-called automatic choice based on the root mean squared errors (RMSE) of retrospective forecasts. We study the sensitivity of the RMSE and inves-tigate the reliability of the automatic choice of parameters for forecasting monthly temperature and humidity recorded at three meteorological stations in Oman.