Comparison of statistical techniques for forecasting set 100 index

Abstract

This research is the comparative study of three forecasting methods on SET 100 index. These methods are Simple Exponential Smoothing method, Holt's method and Box-Jenkins method. Two sets of data were used in this study. The first one used as a training set was collected from January 2, 2008 to September 30, 2011. The other one used as a testing set was collected from October 3, 2011 to October 14, 2011. We propose that the most appropriate forecasting method is the one which give the lowest MAE, MSE, MAPE and diagnostic checking. The results show that Holt’s method displays the lowest MAE, MSE, and MAPE. The next one is Box-Jenkins method while the Simple Exponential Smoothing method exhibits the highest value of all three indicators. However, the model from Box-Jenkins method is the only one way that reveals no residual autocorrelation. This leads to the conclusion that the appropriate models for forecasting of SET 100 Index is ARIMA (2,1,3).