Evaluation of long short-term memory versus recurrent neural networks for demand forecasting in supply chains

Abstract

Abstract Supply chain management is used in both the private and public sectors to meet the needs of consumers while maximizing the economic profitability of an organization. However, in the public sector, military logistics focuses on combat readiness, unlike the private sector, which seeks to maximize profits. Therefore, it is necessary to maintain an appropriate supply level to keep state-of-the-art combat weapon systems operational. Spare parts comprise a large proportion of the equipment maintenance costs. Therefore, several companies have consistently attempted to strike a balance between the cost and supply of spare parts by improving the demand forecasting accuracy. Because the demand for spare parts fluctuates based on equipment requirements, accurate demand forecasting is critical. Fortunately, the recent accumulation of maintenance data has led to more opportunities to improve the demand forecasting accuracy.