Dependence of prognosis and filtration failure on different values of parameters for some classes of non-stationary random sequences

Abstract

The article continues the study of estimates of random functions at a future moment of time, linear with respect to the values of pre-histories of processes. The article considers the dependence of the mean square of the forecast error of a random sequence on the last value at different values of the parameters. For non-stationary random sequences, even with correlation functions of the simplest form, such studies haven’t been conducted. To obtain representations of correlation functions, a Hilbert approach is used to calculate correlation functions as scalar products in the corresponding Hilbert space. Investigations of the dependence of the mean square of the prediction error of a random sequence on the last value at various values of the parameters discussed in the article can be used to simulate filtration and prognosis processes in real systems in the case of non-stationary random signals.