ai-powered selection and classification of resilient suppliers: a hybrid approach using fuzzy dea and ml techniques and its application in the textile industry

A resilient supplier is able to persevere in the face of disruptions and risks. selecting resilient suppliers is crucial for businesses to receive high-quality services quickly and at a low cost. clustering resilient suppliers facilitates identifying the most efficient and resilient ones. mathematical models used to evaluate supplier resilience and cluster suppliers have limitations when addressing large-scale problems and fuzzy data. new techniques, such as machine learning (ml) methods, can be used to mitigate these limitations and predict supplier performance accurately. few studies have used ml methods to cluster suppliers based on resilience criteria in imprecise data environments. to bridge this gap, this study proposes an integrated approach using fuzzy data envelopment analysis (fdea) and ml methods to predict efficiency scores and classify suppliers based on resilience criteria. these methods were applied to evaluate a spinning and weaving factory as a real-life case study, based on resilience criteria. the results demonstrated that among five algorithms- decision tree (dt), random forests (rf), support vector regression (svr), k-nearest neighbors (knn), and logistic regression (lr) - the svr algorithm had the best performance in predicting the efficiency and resilience of suppliers with the accuracy value of .85. additionally, the suppliers were classified into weak, medium, and strong classes. five ml algorithms were used to predict the class of new suppliers. among the lr, dt, rf, knn, and svr algorithms, the dt had the highest accuracy value of 1, while the knn had the lowest accuracy value of .55.