The objectives of this research are to study the movement and construct the most suitable forecasting
model for the monthly condoms export volumes of Thailand on the COVID-19 situation. The data in this study
gathered from the Thai Customs that recorded in monthly during January 2015 to March 2021 a total of 75 months.
Then the data were classified into two sets. The training data set, January 2015 to December 2019, for 60 months
were used to construct model by Decomposition, Holt-Winters Exponential Smoothing, Box-Jenkins and Combined
method. The test data set, January 2020 to March 2021, a total of 15 months were used to compare forecasting
accuracy model the earlier three models via the criteria of Root Mean Square Error (RMSE) and Mean Absolute
Percentage Error (MAPE). Research results indicated that the combined forecasting method where the weighted are
using a regression analysis method has given the lowest RMSE and MAPE, therefore this method is the most suitable
method of monthly export condoms quantity. The forecasting model is:
t 1t 2t 3t
YˆC = −10,200+0.584Yˆ +0.597Yˆ −0.198Yˆ , where 1t 2t
ˆˆY , Y , and 3t
ˆ Y represent the single forecast at time
t from Decomposition, Holt-Winters Exponential Smoothing and Box-Jenkins method, respectively.
Keywords
Decomposition Holt-Winters Exponential Smoothing Box-Jenkins Root Mean Square Error Mean Absolute Percentage Error