.png){kind=icon}
.png){kind=icon}
Return
Metropolitan city supply chain network design under consumer behavior and partial disruptions: A machine learning approach
.png){kind=icon}
PRE
AI
M
M
A
DOI：10.1016/j.cor.2025.107367.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Designing a supply chain network in the post-pandemic era is crucial in metropolitan cities, as recent events have shown how sudden disruptions can severely affect the flow of materials and services. This study proposes a hybrid framework that integrates machine learning and optimization to address a stochastic supply chain network design problem, incorporating customer behavior and potential disruptions. Specifically, unsupervised machine learning techniques are employed for demand clustering, enabling the identification of distinct customer segments based on purchasing patterns. These clusters support more accurate demand forecasting and enable pricing strategies. The clustered demand profiles are then utilized in a Mixed-Integer Nonlinear Programming (MINLP) that optimizes the supply chain network. The model operates under uncertainty, considering stochastic demand and partial disruptions in distribution centers. To manage the risk associated with uncertainties, the Conditional Value-at-Risk (C-VaR) metric is incorporated into the optimization framework. Furthermore, Lagrangian relaxation is used to efficiently handle complex and non-convex constraints. The proposed approach is validated using a real-world industrial case study, demonstrating both its practical relevance and strategic value. Results from a real-world industrial case show that the data-driven clustering improved profit margins by approximately 28% compared to a deterministic benchmark, demonstrating significant pricing effectiveness. Furthermore, the integrated framework achieved a 92% disruption recovery rate and reduced lost sales by over 30% compared to traditional models, confirming enhanced network resilience. Overall, this methodology supports more robust, data-driven decision-making in dynamic and uncertain supply chain environments.
Journal
C
IF:
4.3
Papers:
6.5K
Citations:
1.8W
Organization
Cited Papers