AI-Driven Credit Risk Models for Small-Scale Lending: A Business Analytics Framework for Predictive Performance and Responsible Deployment

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Quantitative analysis compares machine learning models for predicting SME credit risk, indicating improved accuracy and fairness.

The MLP neural network achieved the highest accuracy of 95%, effectively capturing borrower behavior patterns while maintaining transparency.
XGBoost delivered an 86% accuracy but struggled with identifying actual defaulters, revealing a common challenge in imbalanced datasets.
Random Forest demonstrated better recall at 68%, but its overall accuracy fell to 65%, highlighting the trade-offs in model performance.
This research suggests that advanced machine learning combined with interpretability tools enhances credit risk assessments for SMEs.

Oyeyemi et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark