A Suitable Forecasting Model for Export Value of Small and Medium Enterprises in Thailand

Abstract

Background and Objectives: Small and Medium Enterprises (SMEs) constitute an important driving force of Thailand’s economy, playing a vital role in employment generation, income distribution, national economic development, and enhancing the country’s competitiveness. Over the past several years, the export value of Thailand’s SMEs has faced severe volatility driven by both domestic and international factors, including the global economic slowdown, foreign exchange rate fluctuations, trade wars, and the impacts of the Coronavirus Disease 2019 (COVID-19) pandemic. Consequently, these factors have led to high uncertainty in SME export values, directly impacting the financial stability and long-term sustainable growth of Thai SME entrepreneurs. Accurate forecasting data supports production planning, resource management, marketing strategy development, and the formulation of policies to support small and medium enterprises (SMEs) in alignment with changing economic conditions.Therefore, the researcher was interested in conducting a study titled “ A Suitable Forecasting Model for Export Value of Small and Medium Enterprises in Thailand.” The objective of this study was to determine the most suitable forecasting model for predicting the export value of Small and Medium Enterprises (SMEs) in Thailand by comparing forecasting techniques: Holt’s exponential smoothing method, Winter’s exponential smoothing method, and the Box-Jenkins method. The resulting forecasts were intended to provide reliable information for production planning and export management for SMEs, enabling them to align their business operations with changing trends in international trade.

Methodology: The data used in this study consisted of the monthly export value of Small and Medium Enterprises (SMEs) in Thailand, measured in millions of Baht, covering the period from January 2016 to December 2025 (120 months). The dataset was divided into two subsets. The first subset, comprising 108 months of data from January 2016 to December 2024, served as the training dataset to construct forecasting models using time series analysis techniques, specifically Holt’s exponential smoothing method, Winter’s exponential smoothing method, and the Box-Jenkins method. The second subset, consisting of 12 months of data from  January 2025 to December 2025, was used as the testing dataset to compare the forecasting accuracy of the methods. The evaluation was based on the lowest Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) criteria.

Main Results: In this study on the export value of Small and Medium Enterprises (SMEs) in Thailand, the first dataset from January 2016 to December 2024 was used to construct forecasting models. The results were as follows: 1. Holt’s exponential smoothing method, the optimal smoothing parameters were  = 0.301 and = 0.000047     2. Winter’s additive exponential  smoothing, the optimal smoothing parameters were  = 0.409 , = 0.0000021 and = 0.000015  3. Winter’s multiplicative exponential smoothing, the optimal smoothing parameters were = 0.395 ,  = 0.001 and = 0.217  4. For the Box-Jenkins method, the most suitable model was identified as SARIMA(2,1,0)(0,1,1)₁₂ . The second dataset, spanning from January 2025 to December 2025 was used to compare the accuracy of the forecasting methods. The evaluation results were as follows: 1. Holt’s exponential smoothing method yielded an RMSE of 58,634.33 and a MAPE of 4.52% 2. Winter’s additive exponential  smoothing yielded an RMSE of  49,628.21 and a MAPE of 4.33% 3. Winter’s multiplicative exponential smoothing yielded an RMSE of 51,212.36 and a MAPE of 4.43% and 4. the Box-Jenkins method with the SARIMA(2,1,0)(0,1,1)₁₂ model yielded an RMSE of 44,696.63 and a MAPE of 3.90%. Therefore, the Box-Jenkins method provide the highest forecasting accuracy among the methods considered.

Conclusions: This study constructed and compared forecasting models for the export value of Small and Medium Enterprises (SMEs) in Thailand using time series analysis techniques, specifically Holt’s exponential smoothing method, Winter’s exponential smoothing method, and the Box-Jenkins method. Based on the lowest Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) criteria, the results indicated that the Box-Jenkins method with the SARIMA(2,1,0)(0,1,1)₁₂  model provided the highest accuracy, yielding the lowest forecasting errors.