Μ-modification of the “pyramidal” method of data extrapolation
Keywords:extrapolation; forecast; divided differences; interpolation, time series recognition problems
In this work certain generalizations of the classical derivative of a differentiable function are introduced. The method of time series extrapolation is proposed on the basis of corresponding generalizations. In the basis of this method is the analysis of separated differences. It is proposed procedure of the corresponding differences modification and finding of such order, for which it is possible to find in the certain sense the best forecast value. Then the value of the output function at a point that lies outside the interpolation interval is based on the found predictive value for the separated differences using a special computational procedure.