Abstract
Supply lead time constitutes a core parameter in inventory management and plays a critical role in supply chain performance. Yet, how to promote better supply lead time estimations that account for multivariate effects of historical supplier dynamics remains poorly understood. This paper proposes a decision support system that uses a supervised regression strategy with multivariate information for estimating supply lead times. We combine ideas from big data analytics and data mining to explore the effects of different supply-related variables on the dynamics of supply lead time. We design a robust rolling window evaluation scheme to compare both the statistical and inventory performance of different well-known data mining models. Numerical tests with empirical data from a large automotive manufacturer demonstrate that the Random Forest model consistently outperforms other competing models, leading to median decreases of 18%–24% in the mean absolute errors of supply lead time estimations. As a consequence of our results, we also provide insights on how these estimations contribute to the proactive management of safety stocks.
Original language | English |
---|---|
Article number | 106671 |
Number of pages | 16 |
Journal | Engineering Applications of Artificial Intelligence |
Volume | 125 |
DOIs | |
Publication status | Published - Oct 2023 |
Keywords
- Big data
- Data mining
- Lead time uncertainty
- Safety stock
- Supply chain risks