Abed S. Mathematical Models of Dynamic Series Forecasting in Dealing Information Systems

The Thesis is dedicated to elaboration of elements of mathematical software for dealing information systems realizing the procedure of forecasting models construction based on analysis of statistic properties of dynamic series. Consecutive procedure for multi-dimensional models of trend building for stochastic temporary series of foreign currency exchange rates has been improved based on fundamental factors, in particular, correct using of robust and shifted methods of estimation. For the first time, mathematical models of dynamic series forecasting with the consideration of the error forecasting have been developed. Mathematical forecasting model of a forecast error has been synthesized based on the research of exponential alignment methods and their modification of autoregress and integrated sliding mean under conditions of the impact of disturbing factors that allowed to increase the quality of forecasting models. For the first time, mathematical forecasting models have been obtained for foreign exchange currency rates for increasing and declining portions of the temporary series with the assistance of NN, which together with the results of the technical analysis decreased the value of market risks. Scientific results can be applied when developing mathematical software for ACS.