Patcharaporn Saguanchokvanich, Chompoonoot Kasemset, and Trasapong Thaiupathump
Published in Data Science and Engineering (DSE) Record 2025 Vol. 6 No. 1 pp. 433-456
Abstract
Steel sales forecasting is a crucial element in the strategic planning of Chiang Mai Center Steel Co., Ltd. This study focuses on forecasting sales of product WR-44202050, the company’s top-selling item, by comparing various forecasting models, including ARIMAX, LSTM, VARX, and Hybrid ARIMAX– MLP. Model performance was evaluated using Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE). The ARIMAX model achieved the highest accuracy, with MAE of 4.28, MSE of 18.28, RMSE of 4.28, and MAPE of only 0.01%. These results indicate that ARIMAX is the most suitable model for steel sales forecasting in this context, offering strong potential as a decision-support tool for production planning, inventory management, and strategic business operations.