Service Queue Optimization using Decision Tree and XGBoost Based on AHP Weighted Criteria

Abstract

The increasing use of electric vehicles, especially electric motorcycles produced by PT Hartono Istana Teknologi (Polytron), demands a better service queue management system. The current First-Come, First-Served (FCFS) method is considered inadequate because it ignores the level of service urgency and vehicle damage characteristics. This study proposes a priority-based queue system by combining the Analytical Hierarchy Process (AHP) method and machine learning classification algorithms, namely Decision Tree and XGBoost. Data were collected through interviews with workshop technicians, then AHP was used to determine the weights of three main criteria which were then used as features in the service priority level classification process. The evaluation results showed that both models were able to classify priority services well, but XGBoost showed superior and more stable performance than Decision Tree. Quantitatively, Decision Tree achieved an accuracy of 94.44% with an average prediction time of 0.00162 seconds, while XGBoost achieved a higher accuracy of 97.22% with an average prediction time of 0.00202 seconds. Thus, even though it requires slightly longer computation time, XGBoost is considered more suitable for application in service priority systems because it is able to provide more consistent and reliable classification results.