Using Artificial Neural Networks to Forecast Stock Market Indices

Forecasting of financial time series by neural networks requires special preparation of data, selection of size and architecture of neural network, choice of teaching method. The paper shows how the transformation of time series affects the performance of a neural network model on the example of forecasting stock indices UX and PFTS for the Ukrainian stock market. Neural networks were constructed, UX and PFTS stock indices were forecasted using MLP (multilayer perceptron) and RBF (radial basis function) networks. It is shown that the values predicted using time series with lag of input vector is much more accurate.