Smart Sustainability: A Theory-Guided Convolutional Neural Networks With Gazelle Optimization Framework for Supplier Classification

The urgent requirement for resilient and sustainable supply chains has driven the significance of smart supplier selection. As organizations face greater risks of disruption, traditional multi-criteria decision-making methods often lack scalability and adaptability in dynamic environments. To overcome these shortcomings, this work introduces a novel hybrid framework by integrating SCOR 4.0 with Theory-guided CNN optimized using the Gazelle Optimization Algorithm (GOA), providing a unique combination of structured decision models with deep learning for sustainable supplier classification. The SCOR 4.0 system offers a systematic, hierarchical foundation for the assessment of supplier performance in terms of reliability, responsiveness, cost, digital preparedness, and service dimensions. The Best Worst Method (BWM) is utilized to obtain understandable weightings for these measures. CNNs are used to capture intricate, nonlinear relations between criteria, while GOA tunes the hyperparameters of the CNN, yielding higher accuracy and stability of convergence. Experimental outcomes reveal that the suggested CNN-GOA model is capable of acquiring a classification precision of 99.16 % and precision of 93.33 %, which surpasses that of traditional machine learning models. Using normalized acceptance scores helped to rank 28 pharmaceutical suppliers, with S3 and S20 scoring 1.0 at the top and S21 ranking as the lowest at $\mathbf{0}$. Besides strengthening decision-making, the model builds a flexible system for finding sustainable suppliers that is responsive to on-going changes. By combining theory-based deep learning with biologically inspired optimization, we can create smarter and data-driven approaches to supply chain procurement in Industry 4.0.