| dc.description.abstract | This study investigates the optimization of the supply chain at Air Liquide Vietnam (ALV)
– an industrial gases manufacturing corporation, focusing on enhancing demand forecasting
and solving the inventory routing problem (IRP). The motivation stems from the significant
challenges the company faces, highlighting the need for accurate forecasting and efficient
inventory routing to improve operational efficiency and customer satisfaction. Six
methodologies including traditional time serie techniques (ARIMA, SARIMA) and
machine learning – based techniques (RF, XGBoost) and deep learning techniques (LSTM,
ANN) are employed then identified as the best model for demand forecasting process at Air
Liquide Vietnam based on MAPE, MAE, RMSE error measures. Random Forest
outperformed with the MAPE error 3%. Following that, the forecasted demand is monitored
in a fifteen days horizontal and applied to solve IRP problems through a mathematical
model and an iterated local-search (ILS) heuristic which reduce nearly 76% and 26% in
logistics ratio for operation performance.
The results demonstrate a notable increase in forecast accuracy and potential logistics and
inventory cost savings. This study not only provides Air Liquide Vietnam with actionable
insights for overcoming its current logistical hurdles but also contributes to the broader
discourse on supply chain optimization within the air industry. | en_US |
