Predicting the sustainability of supply chains by integrating a novel network DEA model with ML techniques

A supply chain includes multiple interconnected sub-processes through internal activities and operations. This paper proposes a new network Data Envelopment Analysis (DEA) method to assess the performance and sustainability of supply chains taking into account their interconnected members. Additionally, the interconnections among sub-processes and data processing impose numerous variables and constraints on performance evaluation models. State-of-the-art techniques, specifically Machine Learning (ML) algorithms, can be employed in conjunction with performance evaluation methods to address the computational volume and improve prediction accuracy. However, few studies have integrated ML methods with network DEA to evaluate sustainability of supply chains. To address this gap, this paper presents a combination of a novel Network DEA (NDEA) method with ML algorithms to predict the performance of supply chains. The proposed approach was applied to evaluate the tomato paste supply chains of a main company in Iran. The results showed that among three ML algorithm.- Linear Discriminant Analysis, Multi-Layer Perceptron, and Logistic Regressio.- the Multi-Layer Perceptron algorithm had the highest accuracy in predicting efficiency scores, with a rate of about 94%. The Multi-Layer Perceptron algorithm was applied to predict the sustainability of a new supply chain in terms of economic, environmental and social aspects.