Demkivs'ka T. Methods and means of prediction of ecological and economical proceses on the based on time series computer simulation

Working up problems of the elements of information technologies of economic and ecologic systems forecast are solving in the dissertation. ARIMA-models have received a common recognition as the adequate mathematical approach for the description of many natural and economic processes, which are described by scalar and vector time series. The various approaches to automation of ARIMA-models of information streamchoice class and structure are investigated. The expediency of particular correlation coefficient use for search of model structure is experimentally confirmed. This gives an opportunity to construct a rather narrow class of models without participation of experts estimation and empirical conclusions. The numerical algorithm and computer program are developed for choice of the model, which adequately describes a time series. The purposeful search of optimization methods, for estimation of parameters of ARIMA-models is carried out. Among the considered algorithms the S2 method, which does not belon g to a class of gradient methods has the highest efficiency. Methods of construction of a forecast used for creation of subsystem of a forecast in the structure of information system of decision support were developed.