This study presents a hybrid data-driven decision-making framework for evaluating and classifying energy projects in the oil and gas sector, integrating expert judgment with machine learning capabilities. The proposed model combines the Fuzzy Best–Worst Method (FBWM) to determine consistent weights for twenty evaluation indicators and the XGBoost algorithm to predict project acceptance outcomes. The indicators are organized into four key dimensions—resilience, sustainability, digital transformation, and general performance—capturing the multifaceted nature of modern energy project assessment. In the first stage, FBWM was applied to derive reliable fuzzy weightings based on expert evaluations, ensuring structured incorporation of domain knowledge into the model. In the second stage, the obtained weights were used as input features for an XGBoost classifier trained on a dataset of 860 real-world project records. The model achieved strong predictive performance, with an accuracy of 0.95, precision of 0.95, recall of 0.86, and an F1-score of 0.90. Benchmarking against ANN and SVM demonstrated the superior accuracy and stability of XGBoost in handling nonlinear relationships and multivariate project attributes. The findings show that resilience and sustainability indicators—particularly supply chain risk, financial flexibility, carbon emissions, and green ROI—play decisive roles in determining project viability. Moreover, digital readiness aspects, including automation potential and data analytics integration, significantly enhance long-term project success, highlighting the increasing strategic importance of digital transformation in the energy sector. Overall, the hybrid FBWM–XGBoost framework offers a transparent, adaptable, and practical decision-support tool for project screening, risk assessment, and portfolio optimization. It provides energy-sector managers with reliable, data-driven insights that can strengthen resource allocation decisions and improve the resilience and sustainability of project portfolios.
Tahmasebinezhad et al. (Mon,) studied this question.