BAYESIAN MULTIPERIOD FORECASTING FOR ARMA MODEL UNDER JEFFREY'S PRIOR

The main purpose of this study is to find the Bayesian forecast of ARMA model under Jeffrey's prior assumption with quadratic loss function. The point forecast model is to obtained based on the mean of the marginal conditional posterior predictive in mathematical expression. Furthermore, the point forecast model of the Bayesian forecasting compared to the traditional forecasting. The simulation shows that the forecast accuracy of the Bayesian forecasting is better than the traditional forecasting and the descriptive statistics of the Bayesian forecasting are closer to the true value than the traditional forecasting.