A data-driven robust decision-making model for configuring a resilient and responsive relief supply chain under mixed uncertainty

The crucial role of the Relief Supply Chains (RSCs) in the response phase of disaster management is undeniable. However, the literature shows that the simultaneous consideration of the resilience and responsiveness dimensions in designing the RSCs under mixed uncertainty has been ignored by researchers. In this regard, to cover the mentioned gap, the current study aims to configure an RSC by considering two critically important features namely resilience and responsiveness under mixed uncertainty. For this purpose, this work proposed a multi-stage Decision-Making Framework (DMF). In the first stage, a Multi-Objective Model (MOM) is proposed that minimizes the total cost, maximizes the responsiveness level, and maximizes the resilience of the RSC. In the second stage, to deal with mixed uncertainty, a data-driven robust approach based on the Fuzzy Robust Stochastic (FRS), Seasonal Auto-Regressive Integrated Moving Average Exogenous (SARIMAX), and Artificial Neural Networks (ANN) methods is developed. In the third stage, to solve the proposed model, a novel variant of the goal programming method is developed. In general, the main contribution of this study is to develop a novel data-driven DMF to design a resilient-responsive RSC. To show the applicability and efficiency of the developed decision-making method, a real-world case study, the flood that happened in 2019 in Golestan province, Iran, is considered. Eventually, sensitivity analysis, managerial insights, and theoretical implications are presented. According to the achieved results, primary suppliers 1, 3, 5, and 7 and also backup supplier 1 are selected. Also, the results demonstrate that distribution centers 1, 2, 3, and 5 are established. Moreover, the optimal utilization of different transportation modes is specified in the achieved results. The outputs demonstrate that the developed data-driven FRS approach has better performance in comparison with the deterministic and traditional FRS models. Besides, the outputs indicate that the developed solution method has better performance in comparison with the traditional approaches.