Forecasting Financial Time Sesries Using Combined ARIMA-ANN Algorithm

Abstract

The research provides forecasting S&P500 index. Since time series of the stock indices contain both linear and non-linear components, therefore, separately linear ARIMA model and nonlinear ANN models cannot give an accurate estimate of such time series. In this regard we proposed an advanced hybrid forecasting model based on the combination of ARIMA and ANN, based on ARIMA's statistical properties. MSE calculations indicated better accuracy of the S&P 500 stock index forecasts obtained using the proposed algorithm.