Default Probability Prediction by Explainable Machine Learning for Small and Medium-Sized Enterprises

Sep 17, 2026

This paper assesses the credit risk of small and medium-sized enterprises (SMEs) by using explainable machine learning across multiple predictive models. Because SMEs are crucial to economic growth and are frequent recipients of lending, evaluating their creditworthiness remains complex and resource-intensive. Building on the growing use of AI in finance for credit risk modeling, the study focuses on three SME groups in the USA that received support during the Covid-19 pandemic through the Payroll Protection Program (PPP).

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages66

Reading Time2 h 46 min

AccessFree to download

Table of Contents

Abstract

Introduction

Literature Review

Methodology

Machine Learning

Models Accuracy

Data

Paycheck Protection Program (PPP)

Results

Dataset Overall View

Data Analytics

Machine Learning Algorithms

Extreme Gradient Boosting (XGB)

K-Nearest Neighbors Algorithm (KNN)

Logistic Regression (LR)

Decision Trees

Random Forest

Support Vector Machine (SVM)

Naïve Bayes

LIME Analytics

Shap analytics by XGBoost as the reference model

Shap analytics by Logistic Regression as the reference model

Actual data investigation

Other findings

Research Limitations

Conclusion

References

Appendix

Appendix A

Appendix B

Appendix C

Appendix D

Appendix E

Default Probability Prediction by Explainable Machine Learning for Small and Medium-Sized Enterprises Aalborg University Business School Economic and Business Administration Master of Finance 4th Semester Due to: 3rd June 2024 Supervisor Professor Cesario Mateus Student Mehdi Shadpour 20221764 Number of pages: 55 Number of stokes: 85330 Thanks to: The entire faculty of finance at AAU especially my supervisor Prof. Mateus whose support made this research to be done and Joseph & Patryk for standing alongside, working incessantly and for what we learned and experienced together Table of Contents Abstract ........................................................................................................................... 1 Introduction .....................................................................................................................2 Literature Review ..............................................................................................................4 Methodology ....................................................................................................................8 Machine Learning ........................................................................................................... 9 Models Accuracy ......................................................................................................... 11 Data ............................................................................................................................... 11 Paycheck Protection Program (PPP) .............................................................................. 12 Results ........................................................................................................................... 15 Dataset Overall View .................................................................................................... 15 Data Analytics ................................................................................................................ 17 Machine Learning Algorithms ........................................................................................ 17 Extreme Gradient Boosting (XGB) ................................................................................................................... 17 K-Nearest Neighbors Algorithm (KNN) ............................................................................................................20 Logistic Regression (LR) .................................................................................................................................21 Decision Trees ............................................................................................................................................... 23 Random Forest .............................................................................................................................................. 24 Support Vector Machine (SVM) ....................................................................................................................... 25 Naïve Bayes ................................................................................................................................................... 26 LIME Analytics ............................................................................................................. 27 Shap analytics by XGBoost as the reference model ......................................................... 34 Shap analytics by Logistic Regression as the reference model .........................................42 Actual data investigation ................................................................................................ 47 Other findings ................................................................................................................. 52 Research Limitations ..............................................................

FAQs

What problem does the paper address?

The paper targets credit risk assessment for small and medium-sized enterprises by predicting each SME’s default probability.

Which data source is used for the analysis?

It examines three SME groups in the USA that received support during the Covid-19 pandemic via the Payroll Protection Program (PPP).

How does the study ensure interpretability of the machine learning results?

It combines machine learning models with explainable AI methods, including LIME analytics and SHAP analytics using XGBoost and Logistic Regression as reference models.

Sep 17, 2026

Upload by Theodore