FORECASTING TIME SERIES WITH NOTICEABLE FRACTALITY

Dmitri TERZI Universitatea de Stat din Moldova

Abstract

The article considers the main stages of forecasting by neural networks, opportunities and benefits of their use for prediction of time series with a visible fractality. In the examples predictability is determined by the R/S-analysis and can be used to select an appropriate prediction method, including the method of neural networks. Keywords: artificial neural networks, methods of forecasting, R/S analysis, predictability, training samples, sliding window, persistent value.