Machine learning based approach for demand forecasting anti-aircraft missiles

摘要

The Korean military's demand forecasting of spare parts is a major topic for logistics assistance, which considerably affects efficient management of the national defense budget. A time series forecasting method has previously been used, but the results lack accuracy and need improvement. In this study, we collected mass data including 17,451,247 structured and unstructured data on anti-aircraft missile data in the army's DELIIS. Then, we used an ensemble technique to reduce the model's variability and demonstrated increased accuracy for demand forecasting compared with the previous time series method.