Prediction

Abstract

We have shown how SSA can be used to filter a time series to retain desired modes of variability and further how to use SSA to extract a nonlinear trend. Here we discuss how the predictability of a system can be improved by forecasting the important oscillations in a time series taken from the system. The general idea is to filter the record first and then use some time-series model to forecast on the filtered series. There are a couple of time-series models for prediction to choose from. We first present the overall prediction strategy with reference to an autoregressive (AR) model. Then we demonstrate a prediction algorithm that does not require an underlying model.