SYSTEMATIC APPROACH TO DYNAMICS FORECASTING TIME SERIES

Abstract

Time series forecasting is one of the most popular approaches to forecasting the development of economic processes, trade volumes, production and accumulation of products in warehouses, evaluation of alternative economic strategies, budgeting of enterprises and the state, forecasting and management of economic and financial risks, forecasting energy consumption and load on power systems, etc. To date, many methods of forecasting have been described in the special literature. The most common among them are the method of group consideration of arguments [1], autoregression (AR), ARKS, autoregression with integrated moving average (ARIX), autoregression with fractional-integrated moving average (ARDIX), linear and nonlinear multiple 4,5], quantile regression [7], regression trees, neural networks, Bayesian networks, fuzzy sets, fuzzy neural networks and others. However, there is no systematic approach to the choice of mathematical models and methods for forecasting, as well as recommendations for their application.