You have to construct your model first, based on the variables, ... indeed there are many types of models, VAR, VECM, GARCH, SARIMA ...etc, after that the forecasting and tests of evaluation is very easy in Eviews, ref mentioned by Mounir will help you, or you can go to : http://www.eviews.com/Learning/forecasting.html
Thanks you all for your inputs. However, the reference which are given in eviews as attached above, I note that it is forecasting only single equations. I am interested in forecasting my endogenous variables in a VECM specification.
If you want to forecast your endogenous variables using the VECM model, build your VAR model first and test it for autocorrelation, heteroskedastacity, and multivariate normality. When you are satisfied with your results, test the model for cointegartion. Once you find co-integration, you should choose your model using the 'Pantula principle'. Test the long run equilibrium equation for appropriate sign restrictions of the endogenous variables. If the long run equilibrium equation passes the sign restrictions, test the equation for short run 'adjustment parameters'. When you're satisfied with the tests, choose your final VECM equation for forecast purpose and test it for multivariate normality, as well as autocorrealtion and heteroskedastacity. Extend your dataset for the baseline forecast and use the VECM equation for 'in sample' and 'out of sample dynamic forecasting'. Use the 'Root Mean Square' (RMSE) test for alternative model equations and forecast using the best possible VECM model.
Hi Kushneel, I think I've already specified the steps that you should follow forecasting your endogenous model variables. EViews manual can help you more in this regard.