Economical forecasting by exogenous variables

Political and social issues play a big role in economical systems. Macroeconomic variables which are affected by the above mentioned factors can be used in economical forecasting. Time series are used as a very powerful tool in economical systems for short time predicting. As time series predict the future output according to the past behaviors of the system, therefore they can not sense sudden changes in the behavior of the economical system. In this paper, macroeconomic variables are used as exogenous variables in forecasting model. Traditional methods using transformation and differentiation suffer from a decrease in accuracy forecasting. To get rid of the problems in the above mentioned methods, a neuro-fuzzy (NF) structure is used as a strong nonlinear mapping tool even on nonstationary time series. Combination of statistical methods on time series and other dynamical models with NF structure, provide a better model in forecasting. Using ldquoNF-ARMAXrdquo,rdquoNN-ARMAXrdquo models and implementing them on real-life data of ldquoTehran Stock Marketrdquo show a good accuracy in our new designed predictive model.