The electronic payment industry, as one of the essential pillars of digital economy, operates in a dynamic environment characterized by economic fluctuations, rapid technological developments, operational disruptions, and growing service expectations. Under such conditions, selecting and continuously evaluating capable suppliers has become a strategic concern for payment service providers. This study aims to develop an integrated data-driven framework for evaluating agile and resilient suppliers within the context of digital transformation. First, the relevant supplier evaluation indicators were identified and organized into four main dimensions: agility, resilience, digital transformation, and general performance criteria. The fuzzy best–worst method (FBWM), based on the judgments of 18 experts, was then employed to determine the relative importance of indicators under uncertainty. The results indicated that fast delivery, source diversification, safety stock level, and crisis management experience were the most influential indicators, highlighting the critical roles of agility and resilience in the electronic payment industry. In the next stage, the weights obtained from FBWM were incorporated into the feature-scoring and selection process of the CatBoost algorithm. This mechanism allowed expert knowledge to guide the learning process while preserving the model’s ability to discover nonlinear patterns and interactions from actual supplier performance data. A case study was conducted in one of Iran’s leading electronic payment companies to examine the applicability and predictive performance of the proposed framework. The weighted CatBoost model achieved an MAE of 0.3468, an RMSE of 0.4462, and an R² value of 0.8673. The results demonstrated that the weighted model outperformed both the unweighted CatBoost model and the other machine-learning algorithms considered in the comparative analysis. Moreover, comparing the weighted and unweighted versions of the models confirmed that incorporating FBWM-derived weights provided an independent contribution to predictive improvement rather than merely reflecting the inherent performance of CatBoost. The proposed framework integrates expert-based prioritization with data-driven prediction and offers a reliable decision-support tool for supplier selection, ranking, and continuous monitoring. It can also assist managers in identifying strategically important supplier capabilities, allocating procurement resources more effectively, and improving supply continuity in technology-intensive and disruption-prone business environments.