University of TehranJournal of Economic Research (Tahghighat- E- Eghtesadi)0039-896938220031222--10104FAJournal Article19700101In this research, combined forecasting is considered. This model is a new approach that is used in two recent decades and indicates considerable reduction of error in forecasted numbers. In this study, at first, forecasting was done with some different methods that named individual methods. These models consist of exponential smoothing methods, trend analysis, and box-Jenkins. Causal analysis and neural network model. Results of these individual forecasting methods (some selected model) are combined and compared with artificial neural network and multiple regression models. Used data consist of OPEC oil demand from 1960 to 2002 as dependent variable and price, GDP, other energy demand, population, added value in industry as independent variables. In mono-variable methods only dependent variable is entered. Data of 1960-1996 are used for all variables and testing. Data is put under observation between 1996 to 2002. Computed MSE, MAPE, GAPE indexes is shown considerable reduction in errors of forecasting. Keyword: Macroeconomic Forecasting, Time Series, Neural Network, Mix Forecasting.In this research, combined forecasting is considered. This model is a new approach that is used in two recent decades and indicates considerable reduction of error in forecasted numbers. In this study, at first, forecasting was done with some different methods that named individual methods. These models consist of exponential smoothing methods, trend analysis, and box-Jenkins. Causal analysis and neural network model. Results of these individual forecasting methods (some selected model) are combined and compared with artificial neural network and multiple regression models. Used data consist of OPEC oil demand from 1960 to 2002 as dependent variable and price, GDP, other energy demand, population, added value in industry as independent variables. In mono-variable methods only dependent variable is entered. Data of 1960-1996 are used for all variables and testing. Data is put under observation between 1996 to 2002. Computed MSE, MAPE, GAPE indexes is shown considerable reduction in errors of forecasting. Keyword: Macroeconomic Forecasting, Time Series, Neural Network, Mix Forecasting.https://jte.ut.ac.ir/article_10104_bf07bcc73da0f8c11a30b2eeee401424.pdf