A hybrid approach using Z-number DEA model and Artificial Neural Network for Resilient supplier Selection

Abstract

Today's business environment has created a high level of uncertainty and disturbed procedures in supply chains. Suppliers have been often identified as the main source of risks in creating the massive levels of disruptions in supply chains. That is why resilient supplier selection can greatly reduce purchase costs and time delays and can create stability in business practices, thereby increasing competitiveness and customer satisfaction. Pharmaceutical companies play an important key role in the health of society, and these companies are frequently exposed to this disorder. Hence, this paper tries to propose a new integrated approach based on traditional (delivery, quality, price, technology level) and resilient criteria for supplier selection in pharmaceutical companies using the Z-number data envelopment analysis (Z-DEA) model and artificial neural network (ANN). In the proposed approach, expert opinions have been provided based on Z-numbers due to the inherent ambiguity and uncertainty in the evaluation process. This is the first study that evaluates the pharmaceutical industry based on traditional and resilience factors by presenting a methodological structure under the uncertainty environment. Here, a fuzzy mathematical model is used. A real case study is utilized to indicate the applicability of the proposed approach to resilient supplier selection in the pharmaceutical industry. Finally, the suppliers are ranked and the best supplier is selected regarding the reliable level of α. To indicate the features and capabilities of the selected approach, the performance analysis is presented in three parts. First, the obtained results are compared with a fuzzy DEA (FDEA) method in the form of validation and verification. Second, a sensitivity analysis is executed to show the effects of different criteria on ranking results, and the price index is identified as the most important evaluation criteria. Third, a predictive model is presented based on ANN that is able to detect the efficiency or inefficiency of suppliers with an 83% accuracy.