Evaluating the efficiency of banking institutions is essential for understanding their operational effectiveness and competitiveness within the financial sector. Data Envelopment Analysis (DEA) serves as a non-parametric method to assess the relative efficiency of decision-making units based on multiple inputs and outputs, with the underlying production process considered as a black box. Complementarily, Machine Learning (ML) models offer predictive capabilities to forecast efficiency scores based on various financial and operational metrics, while Explainable AI (XAI) techniques further enhance the transparency and interpretability of the models. This study leverages a simulated dataset to integrate ML and XAI techniques with DEA, providing a robust framework that delves deeper into the structure of the black box for sustainable efficiency analysis and prediction in the banking industry.