Machine learning for dynamic multi-product supply chain formation

Recent trend in eCommerce applications toward effectively reducing supply chain costs—including spatial, temporal, and monetary resources—has spurred interest among researchers as well as practitioners to efficiently utilize supply chains. One of the least studied of these views is adaptive or dynamic configuration of supply chains. This problem is relatively new since faster communications over the Internet or by any other means and the willingness to utilize it for effective management of supply chains did not exist a few decades ago. The proposed framework addresses the problem of supply chain configuration. We incorporate machine-learning techniques to develop a dynamically configurable supply chain framework, and evaluate its effectiveness with respect to comparable static supply chains. Specifically, we consider the case where several parts go into the production of a product. A single supplier or a combination of suppliers could supply these parts. The proposed framework automatically forms the supply chain dynamically as per the dictates of incoming orders and the constraints from suppliers upstream.