Application of Radial Basis Functions in Neural Networks for Prognosis of Economic Parameters

摘要

The specific architecture of a feedforward neural network is described. Using example of two real time series, namely exchange rates of Ukrainian hryvna to US dollar and that of Ukrainian hryvna to Russian rouble for the second half of 1997, we demonstrate possibility of prognosis using neural network, whose structure contains radial nodes. The results of operation of a model of such architecture are presented and comparison of its possibilities with those of a classic model of multilayer perception (MLP model) is performed.