머신러닝을 이용한 해군 수리부속 수요예측 사례연구

Abstract

The Naval Logistics Command forecasts demand for repair parts by compiling the time series techniques and qualitative judgment of repair parts item managers. However, the time series technique is difficult to forecast temporary and intermittent demand, and there is a problem in which it is difficult for an item manager dealing with more than 4,000 items to accurately grasp the demand information for each item. This paper proposes a demand forecasting method applied with machine learning techniques that have been actively utilized in many fields recently to improve those problems and increase demand forecasting rate. For the purpose of evaluating the proposed method of forecasting demand for repair parts, the accuracy was compared and analyzed by dividing the results of the forecast by item and quantity accuracy by applying the current procedure and method of forecasting demand for repair parts.