Forecasting techniques based on absolute difference for small dataset to predicttheSET Indexin Thailand

Abstract

This research aims to use asimple statistical method asa forecasting model with a small dataset. Absolute difference methods, average absolute difference and minimum absolute difference, were used to adjust the dataset, i.e., the SET Index, before fitting using thefollowing two forecasting models, anautoregressive forecasting model and asimple moving average forecasting model.Then we comparedthe quality of predictions using the mean square error and the mean absolute difference. These showedthat the mean square error of the average absolute difference filtering method were 15.13%, 15.17% and 7.31% less than the original dataset for aone-period autoregressive forecasting model, atwo-period autoregressive forecasting model and a three-period simple moving average forecasting model,respectively.The mean absolute differenceswere 8.36% , 8.39% and 4.10% less than the original dataset for aone-period autoregressive forecasting model, atwo-period autoregressive forecasting model and a three-period simple moving average forecasting model,respectively. The mean square error of the minimum absolute difference filtering method were 66.02%, 58.94% and 16.33% less than the original dataset for aone-period autoregressive forecasting model, atwo-period autoregressive forecasting model and a three-period simple moving average forecasting model,respectively.The mean absolute differenceswere 39.60% , 33.81% and 9.37% less than the original dataset for a one-period autoregressive forecasting model, atwo-period autoregressive forecasting model and a three-period simple moving average forecasting model,respectively.

Keywords

• Average absolute difference
• Forecasting techniques
• Minimum absolute difference
• Simple statistical method
• Small dataset