Sales Time Series Data Analysis Using a Combination of Supervised andUnsupervised Learning for Business Strategy Optimization at PT X

Jurnal Infra
Vol. 14 No. 1 (2026): Jurnal Infra

PT X is an industrial gas company with an extensive operational network, yet it faces challenges in analyzing sales performance and price inconsistencies across regions. The lack of a data-driven analytical system makes it difficult for the company to identify sales trends, product performance, and branch contributions objectively. This study aims to develop a data-driven analytical system to support business decision-making. Sales trend analysis was conducted using SARIMAX, compared with SARIMA and XGBoost, with SARIMAX showing the highest accuracy. Evaluation using MAPE and RMSE indicated error rates of 18% for SARIMA, 14% for XGBoost, and 5% for SARIMAX. Branch clustering was tested using K-Means and GMM, with GMM selected due to faster execution and better clustering quality (Silhouette Score 0.623 vs 0.622). Additionally, Linear Regression and Random Forest were applied for price standardization across branches, with Linear Regression providing more consistent predictions, high R², and relatively low errors. The results demonstrate an effective analytical system that supports more accurate and consistent business decision-making.