Using artificial neural networks in time series modeling and an application

In this thesis, using artificial neural networks in time series modeling is considered. For this purpose, İstanbul Stock Exchange National-100 index is considered as a time series. Firstly, autoregressive integrated moving average (ARIMA) model which is one of Box-Jenkins models is used for forecasting this time series. Forecasting is done after the most suitable model is found. Secondly, the most suitable model is found and forecasting is done for the same time series by using artificial neural networks. STATISTICA package program is used for both methods and results are compared by using mean square error (MSE) performance measure. Finally, it is expressed that forecasting which is doing with linear neural networks finding by using artificial neural networks has better performance than forecasting which is doing with Box-Jenkins ARIMA model.