Forecasting Model for Thailand's Government Revenue During the Recovery of the COVID-19 Situation

The objective of this research is to identify an appropriate forecasting model for government revenue time series data across four sectors in Thailand: the Revenue Department, the Excise Department, the Customs Department, and the other agencies such as the Stat Enterprises and the Treasury Department. The study utilizes publicly available data from the website Ministry of Finance Open Data, covering the period from October 2012 to April 2023, totaling 127 months. The data set is divided into two subsets: a training data set spanning from October 2012 to April 2022 (115 months) and a testing data set from May 2022 to April 2023 (12 months). Five statistical forecasting methods are employed: additive trend-based linear with seasonal decomposition, additive trend-based logarithmic with seasonal decomposition, smoothing with Holt-Winters' additive method, the Box-Jenkins method, and a combined forecasting approach using regression analysis. The performance of these models is evaluated using the Mean Absolute Percentage Error (MAPE) metric applied to the testing data set. The findings of the study reveal that smoothing with Holt-Winters' additive method is the most suitable model for forecasting the Revenue Department, the Excise Department, and the other agencies. Additionally, the Box-Jenkins method with a SARIMA(2,1,0)(0,0,2)12 model is identified as the most appropriate choice for forecasting the Customs Department.