AI-ENABLED PREDICTIVE RISK INTELLIGENCE FOR SAUDI BANKING SYSTEMS: INTEGRATING SHARIAH COMPLIANCE, ADVANCED FINANCIAL ANALYTICS, AND REGULATORY DATA GOVERNANCE

Abstract

The rapid digital transformation of Saudi Arabia’s banking sector has created significant opportunities for intelligent financial risk management while introducing increasingly complex challenges related to credit exposure, liquidity pressure, financial fraud, cybersecurity threats, operational disruption, regulatory compliance, and Shariah governance. This study develops an AI-driven risk assessment model for Saudi banking systems by integrating Shariah-compliant financial analytics with regulatory-aware data governance. The proposed framework addresses the distinctive requirements of banks operating within the Saudi financial ecosystem, where advanced risk assessment must support Shariah-compliant financial practices while ensuring transparency, cybersecurity resilience, data protection, and regulatory accountability. A multi-source dataset was constructed using published annual reports of Saudi banks, Saudi Central Bank (SAMA) banking statistics, Tadawul-listed bank disclosures, financial stability indicators, Shariah governance reports, macroeconomic variables, and structured banking-risk records. To enhance the realism and analytical depth of the framework, anonymized and simulated transaction-level data were incorporated, including customer financing behavior, delayed repayment patterns, digital payment activities, suspicious transaction indicators, abnormal login attempts, cybersecurity alerts, and Shariah-compliance risk signals. The collected data were processed through feature normalization, missing-value treatment, class balancing, contract-based feature engineering, risk-category labeling, and regulatory attribute mapping. Multiple machine-learning approaches were evaluated, including logistic regression, support vector machine, random forest, gradient boosting, and the proposed explainable ensemble risk assessment model. Experimental results demonstrate that the proposed model achieved superior predictive performance, attaining 96.4% accuracy, 95.8% precision, 96.1% recall, 96.0% F1-score, and a ROC-AUC of 0.982. Compared with conventional statistical and individual machine-learning models, the proposed framework reduced false risk classifications by 18.7% and enhanced the early identification of high-risk financing cases, suspicious digital transactions, cybersecurity vulnerabilities, and potential Shariah non-compliance events. Explainable AI analysis further identified repayment irregularity, financing concentration, contract type, customer risk history, liquidity stress, abnormal transaction frequency, and cybersecurity risk flags as the most influential risk predictors. The proposed framework provides a Saudi-specific, Shariah-aligned, and regulatory-aware approach for transparent risk scoring, secure data governance, continuous compliance monitoring, and responsible data-driven decision-making in digitally evolving banking environments.