Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
March 12, 2026Open Access

Predictive Modelling of Credit Default Risk Using Machine Learning and Ensemble Techniques

View Full Paper
Ask AI
Bookmark
Share

Authors

MMMofoka Rebuseditsoe MathibelaDMDaniel Maposa

Discussion

Loading...

Member takes

Overview

Demonstrates improved predictive accuracy and interpretability in credit risk modeling using ensemble techniques.

Key Points

  • The aim is to develop a framework that balances predictive accuracy and interpretability in credit default risk models.
  • Utilized the German Credit Dataset for model training and validation.
  • Implemented preprocessing steps including feature encoding, scaling, and SMOTE for class imbalance.
  • Combined four models—logistic regression, Random Forest, XGBoost, and Multilayer Perceptron—using a Stacked Ensemble.
  • Applied SHAP analysis for interpretability of prediction results.
  • Achieved an AUC of 0.761, the highest among all tested models.
  • Precision was 0.783, and recall reached 0.806, contributing to an F1 score of 0.794.
  • Random Forest outperformed XGBoost in terms of AUC despite conventional expectations.
  • Reduced false positives from 39 to 16, enhancing practical applicability of the model.

Cite This Study

Mathibela et al. (2026) studied this question.

synapsesocial.com/papers/69b25afb96eeacc4fcec9422https://doi.org/10.3390/mca31020045
View Full Paper
Ask AI
Bookmark
Share