Prediction of Stock Index Behavior using a Hybrid Model of Artificial Neural Network and Wavelet Transform

Predicting the future values of time series (signals) based on the past and present data can be a useful tool for financial applications. In this study, the stock index prediction is made using a hybrid model, composed of the wavelet transform and Nonlinear Auto-Regressive with eXogenous Inputs (NARX) neural network. This model comprises two stages. First, the preprocessing stage in which the Haar wavelet transform and a soft thresholding approach are used to de-noise the macroeconomic variables (oil prices and Rial to Dollar exchange rate). Second, the NARX neural network is developed to use the de-noised external inputs to predict the stock index as the output. By comparing the results of NARX neural network and the hybrid model, it is shown that the proposed model is more accurate than NARX neural network, and can achieve more reliable results.