Oleshko D. Information technology for speeding-up of the syn-thesis of neural networks for a forecasting problem solving at a decision making

The thesis is devoted to a decreasing of synthesis time for forecasting neural network models and to a rising of a forecast's reliability by intellectualization of training sample forming and of parametric and structural synthesis of neural networks. There are offered in this work: a number of characteristics of training sample; the complex evaluating criterion which have been set up on these characteristics which allows to automize the process of qualitative training sample forming. The approach to formalising of structural synthesis of neural networks is offered. Mechanisms which allow to speed up essentially the process of parametric synthesis at saving quality of tutoring and operation of a neural networks are developed. Also the information technology for synthesis of forecasting models, based on neural networks, for solving tasks in such areas as control, electronic busyness and economic monitoring, is developed in this work.