POLIBITS, Vol. 62, pp. 59-67, 2020.
Abstract: Knowing the demand for products in advance would be ideal for companies that strategically relocate products in their warehouses to facilitate the picking process, which is the most expensive activity in a warehouse. In this sense, an assembly plant from Ciudad Juárez, Chihuahua, Mexico, that handles approximately 10,477 parts in its inventory, periodically relocates them in warehouse zones to facilitate the picking process. However, relocation is done empirically based on the total number of outbound inventory movements of each of the parts made in a given time. This paper describes the implementation of time-series models to forecast the demand for parts that could improve the relocation process. For this purpose, different Holt-Winters Seasonal and SARIMA models were implemented. For the implementation of the SARIMA models, the Box-Jenkins methodology was followed. The AIC and BIC metrics were used to identify the best Holt-Winters Seasonal model and the best SARIMA model. Tests were performed on the residual series to check that model is fit to the data. The RMSE and MAPE metrics were used to evaluate the performance of Holt-Winters Seasonal and SARIMA models. The results of the evaluation carried out indicate that the SARIMA model outperforms to Holt-Winters Seasonal model.
Keywords: Forecast demand, Holt-Winters seasonal model, seasonal auto-regressive integrated moving average model, time series, Box-Jenkins methodology
PDF: Forecasting the Demand of Parts in an Assembly Plant Warehouse using Time-Series Models
PDF: Forecasting the Demand of Parts in an Assembly Plant Warehouse using Time-Series Models
https://doi.org/10.17562/PB-62-7
See table of contents of POLIBITS 62.