On Predictions under Conditions of Uncertainty

A method is created for predicting the trajectory of a difference equation with constant coefficients that is observed under noisy conditions; the method does not require a priori knowledge of the coefficients of the equation. It is shown that the method realizes optimal behavior (in terms of the nature of the dependence on the observation time) of the standard prediction error.