Adaptive Demand Forecasting System for Spare Parts of K-MBT

Purpose: This study was conducted to develop an adaptive demand forecasting system that flexibly reevaluates and applies forecasting models according to demand changes, focusing on high-performing forecasting model groups to enhance the demand forecasting accuracy for the main battle tank (K-MBT).BRMethods: Twenty-one types of time-series models and machine learning/deep learning models were selected for evaluation. Statistical methods were then applied to select a group of 10 high-performing models. Subsequently, the prediction accuracies of all the 21 initially selected models were compared with the results obtained using 10 high-performing models.BRResults: The approach of selecting few high-performing models (10 models) for demand forecasting, compared to using many models (21 models), proved to be superior in terms of accuracy.BRConclusion: For ensuring the operational readiness of weapon systems through the forecast of spare parts demand, we anticipate the continuous enhancement and development of adaptive demand forecasting systems, along with their practical applications.