A hybrid ensemble and AHP approach for resilient supplier selection

Suppliers play a crucial role in achieving the supply chain goals. In the context of risk management, suppliers are the most common source of external risks in modern supply chains. The recognition that continuity of supply chain flow under disruptive event is a critical issue has brought the attention of companies to the selection of resilient suppliers. In contrast to the extensive number of researches on traditional and green criteria of supplier selection, the criteria associated with resilient supplier selection are not well explored yet. This paper first seeks to explore the resilience criteria for supplier selection based on the notion of resilience capacities which can be divided into three categories: absorptive capacity, adaptive capacity, and restorative capacity. Absorptive capacity refers to the capability of system to withstand against disruptive event in prior or called as preparedness of supplier, while adaptive and restoration capacities imply the capability of supplier to adopt itself and restore from disruption or recoverability of supplier. We identified eight effective elements for resilience capacities which contribute to the resilience of suppliers. Advanced data mining approaches like predictive analytics models are used to predict the resilience value of each supplier. We applied ensemble methods by combining binomial logistics regression, classification and regression trees, and neural network to obtain better predictive performance than individual algorithm from the historical data to predict individual supplier's resiliency. This resilience value, obtained from ensemble methods, is coupled with additional four variables to assess the suppliers' overall performance and rank them using different supplier selection models. Finally, a case study has been performed on international plastic raw material suppliers for a U.S. based manufacturer.