← Back to Publications List

Machine Learning Models for Credit Risk Assessment

Students & Supervisors

Student Authors
Afzalul Abid Nazir
Bachelor of Science in Computer Science & Engineering, FACULTY OF SCIENCE & TECHNOLOGY
Islam Saiful
Bachelor of Science in Computer Science & Engineering, FACULTY OF SCIENCE & TECHNOLOGY
Supervisors
Md. Faruk Abdullah Al Sohan
Lecturer, Faculty, FACULTY OF SCIENCE & TECHNOLOGY

Abstract

Financial institutions can minimize losses and make well-informed lending decisions with the help of an accurate assessment of credit risk. This study evaluates various machine learning models, including logistic regression, naive Bayes, random forest, XGBoost, gradient boosting, AdaBoost, Decision Tree, SVM, and KNN, as well as deep learning baselines (LSTM and CNN). Accuracy, precision, recall, F1-score, AUC-ROC, MAE, MSE, RMSE, and MAPE were used to evaluate the models. With an accuracy of 97.18% and an AUC-ROC of 0.9837, Random Forest outperformed the other models. With respective accuracies of 94.61% and 95.04%, XGBoost and Decision Tree also demonstrated strong performance. With an accuracy of 86.79%, Logistic Regression performed worse. Also, deep learning models performed less effectively (80–83% accuracy) on this structured dataset. Random Forest was the most reliable and successful model, providing better predictive power to assess credit risk and assist financial institutions in making informed decisions.

Keywords

Random Forest XGBoost Loan Default Prediction Credit Risk Assessment Financial Risk Management

Publication Details

  • Type of Publication:
  • Conference Name: 4th IEEE International Conference on Signal Processing, Information, Communication and Systems (SPICSCON-2025)
  • Date of Conference: 21/11/2025 - 21/11/2025
  • Venue: University of Rajshahi, Rajshahi-6205, Bangladesh
  • Organizer: IEEE Bangladesh Section and IEEE Signal Processing Society Bangladesh
View on Publisher