Analyzing Fraud Risk Management, Environmental Disclosures and Stewardship in Nigeria's Oil and Gas Sector: A Machine Learning Method Towards Environmental Sustainability

Abstract

This study investigates bidirectional relationships between fraud risk management and environmental stewardship as well as fraud risk management and environmental disclosures through an integrated theoretical framework combining Meta-model fraud theory with stakeholder, agency, and legitimacy theories. Using Random Forest and structural equation modelling, four hypotheses examined the interconnected nature of fraud prevention and environmental disclosure practices. The research analyzed fraud risk management influence on environmental disclosures (H1), environmental disclosure impact on fraud risk management (H2), environmental stewardship effects on fraud risk management (H3), and fraud prevention impact on environmental protection (H4). Advanced machine learning techniques ensured robust statistical validation. The tested hypotheses' results provide empirical evidence supporting these hypotheses, with R2 values ranging from 0.5149 to 0.7719. H1 achieved the highest explanatory power (R2 = 0.7719, 77.19% variance), followed by H2 (R2 ≈ 0.75) and H3 (R2 ≈ 0.73), while H4 showed moderate effects (R2 = 0.5149). Findings indicate that machine learning models demonstrated strong predictive relationships between fraud risk management and environmental outcomes. All hypotheses were supported, showing that fraud management significantly influences environmental disclosures and stewardship. Thus, organizations can leverage fraud controls to enhance environmental performance. In addition, utilizing data-driven approaches enables targeted policy interventions and operational improvements in Nigeria's oil sector. The study contributes to corporate governance theory by demonstrating interconnected risk management and sustainability practices.