A multi-stage machine learning model to design a sustainable-resilient-digitalized pharmaceutical supply chain

Abstract

The significance of the Pharmaceutical Supply Chain (PSC) has been bolded during the COVID-19 pandemic when the demand for pharmaceutical products has drastically increased. The literature shows that the simultaneous consideration of resilience, sustainability, and digitalization in the PSC network design problem, especially using data-driven approaches, has been ignored by previous works. Hence, the current work aims to cover these gaps by proposing a machine learning-based model to design a PSC with resilience, digitalization, and sustainability dimensions. For this purpose, in the first stage, the potential suppliers are assessed using a Random Forest Regressor (RFR). Afterwards, a mathematical model is developed to design the PSC in which the resilience and sustainability aspects are incorporated. Then, a recently introduced method named Fuzzy Lexicographic Multi-Choice Archimedean-Chebyshev Goal Programming (FLMCACGP) is employed to achieve the optimal solution. To represent the application and efficiency of the developed model, a real-world case study in Iran is examined. It should be noted that the demand for products is estimated using the machine learning approach. Overall, the main novelty of this study is to design a sustainable-resilient-digitalized PSC network using a data-driven model. The model identify the most important indicators for the research problem wherein delivery time, quality, backup supplier, robustness, and cost are the most significant indicators. Furthermore, the proposed mathematical model selects the blockchain-based platform to establish the Information-Sharing System (ISS). The effectiveness of the developed methodology is then assessed by comparing its results with the traditional methods. Finally, managerial insights are offered based on the practical implications of the findings.