Intoday's world, competition among firms has evolved into competition among supply chains. Supply chains must be resilient to maintain operational continuity in highly risky, turbulent environments. Suppliers, as the first and outermost layer of the supply chain, are the most vulnerable to risks, and disruptions in supplier performance can affect the entire supply chain. Therefore, incorporating resilience into supplier selection is essential. This study proposes an integrated approach combining Multi-Criteria Decision-Making (MCDM) and machine learning to select resilient suppliers, leveraging both quantitative and qualitative data. First, resilience criteria were identified through a comprehensive literature review, and then localized in collaboration with experts from a domestic steel company. Ultimately, 15 resilience criteria and 17 financial indicators were determined as the final evaluation criteria. To assess suppliers, Shannon entropy and the Measurement Alternatives and Ranking according to Compromise Solution (MARCOS) method were used to weight and calculate resilience scores. These scores, together with financial data, were then used in a machine learning framework comprising Principal Component Analysis (PCA) and K-means clustering. By reducing the data to four principal components, the 24 main suppliers of the case study company were clustered into five groups. Model validity was examined by comparing supplier rankings obtained from the MARCOS and TOPSIS methods, yielding a Spearman correlation coefficient of 0.87, which indicates a strong relationship. In addition, the elbow method and silhouette index confirmed that five clusters were appropriate. By integrating data-driven and knowledge-driven approaches, this research provides a practical step toward improving decision-making in resilient supplier selection.