A two-phase machine learning-based framework for designing viable supply chains: a novel data-driven decision-making approach under incomplete data

What the paper says

Purpose The prevalence of severe disruptive events has emphasized the importance of resiliency and viability in global supply chains. Suppliers, as key sources of external risks, are critical pillars of viable supply chains under disruption propagation (i.e., the ripple effect). This research aims to develop a two-phase supplier selection and order allocation framework to enhance supply chain viability under uncertainty. Design/methodology/approach In the first phase, a novel data-driven decision support system is proposed for evaluating and selecting suitable suppliers under incomplete data. A risk-averse stochastic model is proposed in the following phase to allocate orders to suppliers under disruption propagation. To control the fluctuation of production amount in manufacturing industries over the time horizon, a method is proposed to smooth the production trajectory over the planning horizon, and its effect on remaining parts in stock is investigated. Findings The findings reveal the effectiveness of the proposed framework in overcoming the resilient supplier selection and order allocation problems under incomplete data and uncertainty. The production smoothing method can also reduce production fluctuations over the planning horizon and provide better inventory management. Practical implications Several managerial implications are also included by using the proposed framework in a case study of the home appliance industry. Originality/value This study integrates a novel supplier selection approach under incomplete data, order allocation under disruption propagation, and production smoothing in a viable supply chain framework that has not been investigated previously.