A Strategic Framework for Supplier Performance Assessment and Anomaly Detection in High-Tech Supply Chains: Integrating Shared-Resource DEA With Machine Learning

Evaluating supplier performance in high-tech supply chains is challenging because suppliers must allocate shared resources across operational and sustainability-related activities while maintaining resilience under uncertainty. This study develops an integrated decision-support framework that combines two-stage shared-resource data envelopment analysis (DEA) with machine learning for supplier assessment, anomaly detection, and efficiency prediction. Using NVIDIA's semiconductor supply chain as the empirical setting, the framework explicitly models resource allocation between operational and ESG objectives, thereby addressing the measurement distortion that arises when conventional DEA assumes independent resource use across stages. The proposed approach further incorporates multimodel anomaly detection to identify suppliers with abnormal performance patterns and employs predictive analytics to support forward-looking monitoring. Research rigor is strengthened through transparent data construction, comparative model benchmarking, time ordered validation, and multi-method analytical triangulation. The findings show that the proposed framework produces more reliable and more discriminating supplier assessments than conventional efficiency-evaluation approaches, while also offering earlier identification of performance risks. For engineering management, the framework provides a practical basis for balancing operational and ESG priorities, prioritizing supplier development efforts, and allocating managerial attention to suppliers requiring early intervention.