Использование временны́х рядов для формирования промежутков однотипного поведения параметров объекта при различных способах прогнозирования

Abstract

The new approaches to formation of intervals of the same object parameter behavior on time-seriess in the form of parameter consecutive value sets (coordinates of the "neighbor vectors ") close to a set of their last values used when forecasting. It’s shown that the constructed technique allows to increase number of neighbor vectors to find statistical characteristics used at forecasting. The uniform algorithm for receiving a set of neighbor vectors in each of the considered options is offered. The algorithm of forecasting for one step based on the analysis of coordinates of the neighbor vectors in a special way constructed is developed.