Forecasting the number of foreign workers in Thailand by exponential smoothing method

Abstract

The objective of this research was to study the technique and to compare forecasting methods using the exponential smoothing method which consists of 3 techniques, namely Simple Exponential Smoothing, Trend Method, and Holt-Winter's Seasonal Method. This study collected data on the number of foreigners permitted to work throughout the Kingdom from the statistical database of the Bureau of Foreign Workers Administration, Department of Employment. The data were divided into two sets, the first set from January 2007 to December 2020 is used for predictive model (Training data). As for the second set of data from January 2021 to December 2021 is used for checking the accuracy of the forecast (Testing data). The results showed that the Damped Holt-Winters' Multiplicative Method was the most accurate method for forecasting because it gives the smallest value of RMSE and MAPE value with RMSE of 196760 people and MAPE = 7.769%, followed by Damped Holt-Winter's Additive Method with RMSE of 213,832 people and MAPE of 8.596%. Both methods yielded quite accurate forecasting results.