Comparación de métodos estadísticos y aprendizaje automático para la mejora del proceso de Slotting
Agusto Guillermo Sanchez Ferrer, José Herrera · Revista de investigación de Sistemas e Informática · 2024
Distribution centers or supply warehouses face significant challenges in the order-picking process. A particular issue is the congestion of operators in the picking zone due to the inaccurate distribution of products in limited locations. To address this issue, "Slotting" process has emerged, aiming to efficiently allocate products to specific locations using analytical methods, enhancing order preparation while concurrently reducing operational costs. However, seasonal variations and fluctuations in trends contribute to oscillations in product demand. Therefore, it is suggested to integrate statistical forecasting methods and machine learning to monitor and anticipate changes of products in picking zones. In this context, this research presents two approaches, the first is to propose a new “Slotting” process model based on the literature review and the second is a comparison between methods such as double and triple exponential smoothing; Seasonal Autoregressive Integrated Moving Average; Long Short-Term Memory and Gated Recurrent Unit were proposed to forecast product demand. To validate the methods, a test was carried out with 50 products and a 7-year historical database from January 2016 to May 2023. This research suggests a hybrid approach in which the methods utilized can be a practical strategy to achieve the best forecasting accuracy, optimizing the Slotting process.