As global industrial operations expand, the complexity and volume of suppliers have increased, making sustainable supplier selection (SSS) a strategic imperative for resilient supply chains (SCs). Traditional evaluation methods often fail to handle large datasets and dynamic sustainability metrics, resulting in suboptimal decisions. This study introduces a novel hybrid intelligent framework that integrates Most Productive Scale Size Data Envelopment Analysis (MPSS-DEA) with three Artificial intelligence (AI) algorithms—Artificial Neural Networks (ANNs), K-Nearest Neighbors (KNN), and Chi-squared Automatic Interaction Detection (CHAID)—to enhance both accuracy and scalability in supplier evaluation. Applied to 362 suppliers and 38 sustainability criteria in the petrochemical industry, the proposed framework achieved high classification accuracies of 92.5% (DEA-CHAID), 91.9% (DEANN), and 91.4% (DEA-KNN). The model also demonstrated strong discrimination power, with ROC AUC scores of 0.96, 0.95, and 0.94, respectively. Predictor importance analysis revealed that some of the features, such as R2, SE2, QA4, CA6, and DE3, were the most influential features across all models. Beyond performance metrics, the framework offers real-time supplier replacement, reduced computational complexity, and modular adaptability across industries. It supports ethical and sustainable sourcing by integrating economic, environmental, and social dimensions into decision-making. The intelligent architecture enables lifecycle analysis, promotes transparency, and aligns with global sustainability standards. This research contributes a scalable, interpretable, and data-driven solution for sustainable supplier selection, bridging the gap between traditional DEA models and modern artificial intelligence (AI) techniques. Its applicability across diverse industrial contexts positions it as a robust tool for strategic procurement and supply chain resilience.
