Efficiency Of Estimation And Prediction Of Case-Space Exponential Models And Conventional Methods (Empirical Eomparison)

Abstract

The research aims to build a suitable model for forecasting in the data of the Iraqi economy based on the State Space Model (SSM) with the exponential smoothing methodology. Therefore, two types of prediction methods were addressed, namely, the traditional exponential smoothing methods and the exponential smoothing under State Space Models, which laid its foundation (Hyndman, et al, 2002). The Time Series Automatic Forecast algorithm, which was suggested by Hyndman & Khandakar, 2008, was used as one of the useful statistical tools in forecasting.On the practical side, the time series that represents the annual production index of livestock in Iraq was used, and it includes meat and milk from all sources, and dairy products such as cheese, eggs, honey, raw silk, wool and leather. And then the comparison between the two methods in estimating efficiency using the two criteria (AIC, BIC), and the comparison in the efficiency of predictive performance using standards (RMSE, MAP, MAPE, MASE), within and outside the sample for the method and the model using prediction accuracy measures. Where the time series was broken up and then forecasted outside the training group that constituted 80% of the observations and compared with the test group values. It was concluded that the simple exponential smoothing model with multiplying errors ETS (M, N, N) was the most efficient in estimating and the simple exponential smoothing method (N, N) was the most efficient in performing the prediction. The two researchers recommended that the Ministry of Agriculture and Irrigation should approve the results that were reached in order to formulate a policy for livestock in Iraq.