A machine learning-based hybrid approach for maximizing supply chain reliability in a pharmaceutical supply chain

Abstract

In today's interconnected global economy supply chain (SC) reliability is crucial particularly in sectors like the pharmaceutical industry where disruptions can significantly impact public health. SCs have become important to industries due to a customer-driven shift aimed at improving SC reliability especially in terms of delivery performance. It is crucial to define and find the best strategy for reaching the organizational objectives in SC. While designing a SC supplier selection (SS) and order allocation are two decisions that have to be made separately. This study addresses the critical challenges of SS and order allocation within pharmaceutical SCs. It proposes a novel two-phased hybrid approach the first phase integrates machine learning(ML) and multi-criteria decision-making(MCDM) method for robust SS. The second phase develops a mathematical model to optimize order allocation while considering SC reliability. This work employs support vector machine (SVM) as the particular ML method in which the training data are historical corporate data that dictate parameters weights. These weights are then used in the measurement of alternatives and ranking according to compromise solution (MARCOS) method to rank the suppliers. A multi- objective mixed integer programming (MOMIP) model is then formulated to identify the right order quantity from the identified suppliers of a pharmaceutical SC in order to minimize SC cost and maximize SC reliability. The results indicate that by optimizing SC reliability and costs orders are directed to high-priority suppliers. This study provides a comprehensive data-driven decision-making framework to assure SC's reliability and cost-efficiency. The implications of the findings are also profound and contribute valuable insights for industry practitioners to improve the performance of SC. To illustrate the proposed methodology an SC example of a pharmaceutical industry is analyzed using the LINGO solver. © 2024 Elsevier B.V. All rights reserved.