Tarasenko R. Method of the analysis and improvement of quality of learning set for neural nets in time series forecasting

The object is process of neural nets based time series forecasting; the subject is forming of learning sets (LS) for forecasting; the aims are decreasing error of forecasting and decreasing of the time for forecasting model forming; the methods are theory of probability, statistics, theory of artificial neural net; the new approach lies in criteria of LS quality, scientific explanation of dependencies between LS quality and complexity of basic, for this LS, time series, the technique of improvement of quality of the learning set; the main includes developed software application for LS construction and the preliminary analysis; all this thesis have been applied in Institute of economical and ecological researches of Ukrainian NAS, dealing center "Ostriv" and the Odessa national polytechnic university; the field is decision supporting systems