Fatma M. The investigations of fuzzy logic system in the problems of forecasting and classification

The thesis is dedicated to investigations of fuzzy logic systems and fuzzy neural networks in the problems of forecasting and classification under uncertainty conditions. The new learning algorithm was elaborated and investigated for fuzzy neural networks with Gaussian membership functions. The algorithm of macroeconomic indexes forecasting is suggested with application of fuzzy neural networks. Efficiency analysis of various fuzzy logic inference algorithms for forecasting in macroeconomics and financial sphere was carried out. As a result the best algorithm for this problem was determined. The investigations of rule base size and the number of linguistic values influence on the forecasting accuracy was explored. The learning algorithm of fuzzy classification system was improved and the influence of stepsize on the convergence of learning algorithm was experimentally provedt.