Intelligent Loan Risk Assessment: A Machine Learning Framework for Personalized Credit Evaluation

Abstract-Banks are essential to the global financial system, and one of their primary sources of income comes from loan interest. However, if borrowers fail to repay these loans, it can turn profits into substantial losses, highlighting the importance of assessing the risk of default before approving a loan. Machine learning techniques can be an effective method for quickly and accurately evaluating whether a credit risk should be approved. This study explored six machine learning models—Decision Tree, Random Forest, Support Vector Machine (SVM), Multi-layer Perceptron (MLP) Artificial Neural Network, Naive Bayes, and a stacking ensemble model—to predict the credit risk associated with a loan. Using a dataset of twenty factors typically found in loan applications, the stacking ensemble model achieved the highest accuracy at 78.75%. The Random Forest model, though slightly less accurate at 78.15%, was more efficient while yielding comparable results. Key factors such as credit amount, account status, age, loan duration, and loan purpose were identified as the most influential indicators of credit risk. The findings of this research further support the efficacy of machine learning models for predicting loan default risk.