Prediction of chaotic dynamical processes based on detection of regular component

The problems of detecting components and predicting dynamical processes are considered. Schemes for predicting chaotic time series that are based on detecting their regular, anomalous, and chaotic components followed by applying one of the described prediction methods to the regular component are presented. Regular components are detected using robust linear splines and singular spectrum analysis. Provided examples show that the presented schemes allow predicting dynamical processes with acceptable accuracy.