Application of Machine Learning in Auditing and Credit Risk Assessment: Evidence from Mongolia

In recent years, artificial intelligence (AI) and machine learning (ML) methodologies have rapidly penetrated the fields of auditing, financial analysis, and credit risk assessment, enabling more accurate and real-time evaluations compared to traditional statistical approaches. However, in developing countries such as Mongolia, the integration of these methods into audit and credit evaluation systems remains limited and underexplored. This study aims to develop an integrated model for assessing audit and credit risk and identifying the key influencing factors using machine learning techniques. The analysis is based on data from 88 enterprises that received loans from the Mongolian Small and Medium Enterprise Development Project during 2019–2024, including their financial statements, on- site audit reports, and loan repayment records from the SME Development Fund. Classification algorithms such as Random Forest, Gradient Boosting, and Decision Tree were applied, and their performance was compared using evaluation metrics including Accuracy, Precision, Recall, and F1-score. The results revealed that the Random Forest algorithm achieved the highest performance (Accuracy = 0.944, Recall = 1.000), demonstrating its ability to identify high-risk entities with 100% recall. SHAP analysis indicated that tax arrears, overdue loan days, and non-compliance periods were the most influential variables affecting audit and credit risk. These findings highlight the potential of adopting AI-based integrated risk assessment systems in Mongolia’s auditing and credit supervision sectors, contributing to early risk detection, optimized allocation of supervisory resources, and enhanced transparency at the policy level.