RECONFIGURABLE COMBINED FORECASTS IN A NONSTATIONARY INFLATIONARY ENVIRONMENT

This paper presents a composite method for non-stationary economic for ecasts, incorporating reconfigurable exponential trend extraction and linear combinations of parabolic and spectral estimators for short-ter m prediction. The method automatically identifies the points of time s eries misalignment induced by sharp environmental changes. An applicat ion to the problem of hard currency exchange rate prediction in Russia is presented.