DETERMINATION OF FACTORS AFFECTING PROGRAM SETUP TIMES IN PRODUCTION PLANNING USING MACHINE LEARNING METHODS

In this study, production-related data obtained from the corporate ERP/MES infrastructure (e.g., SAP) of an industrial fastener manufacturing facility operating in İzmir, Turkey, covering the years 2014–2025, were utilized to identify the factors influencing program setup time and to develop a predictive model to support production planning decisions. Program setup time is defined as the duration between the completion of production for the previous part and the point at which the machine becomes ready to resume mass production with the next part. Both statistical methods and machine learning models were employed for this purpose. The dataset used in the study belongs to four press machines of the same model, selected to provide a homogeneous analysis environment due to their processing of similar and identical product groups. Following data preprocessing, descriptive statistics, correlation analyses, regression models, and machine learning approaches (Linear Regression, Random Forest, XGBoost, and Artificial Neural Networks) were applied. Comparison metrics indicated that the Random Forest model achieved the highest predictive performance, followed by XGBoost. In contrast, the lower performance of the linear regression model was attributed to its limited ability to capture nonlinear and interaction-based relationships among variables. The results revealed that variables such as metric change between the previous and the new product, piercing pin diameter difference, transitions in product types, time elapsed since the last production, raw material/diameter relationship, and operator competency exhibited statistically significant effects on program setup time.