💳 Credit Risk Prediction

A Comparative Study of Classification Models and WOE-Based Feature Transformation

This project focuses on predicting whether a client will experience serious delinquency within the next two years using classification models. The dataset was preprocessed to ensure high data quality, including imputation, outlier treatment, and transformation using Weight of Evidence (WOE) to enhance interpretability and model effectiveness—especially for logistic regression.

Three machine learning models were evaluated to identify the best-performing approach in terms of accuracy, recall, and business relevance.

---

🧠 Models Compared

• Logistic Regression (with and without WOE)
• Random Forest Classifier
• XGBoost Classifier

🧩 Feature Strategy

1. Raw Features – Cleaned but untransformed dataset
2. WOE-Transformed Features – Variables transformed using Weight of Evidence to support interpretability and improve logistic regression performance

---

🔍 Key Findings

• Random Forest delivered the best balance of performance metrics:
  • ROC AUC: 0.8569
  • Recall: 0.72
  • Precision: 0.22
• Logistic Regression performed well with WOE-transformed features:
  • Recall: 0.70
  • Precision: 0.19
  • Enabled creation of an interpretable scorecard
• XGBoost had the highest precision for non-delinquents but a low recall (0.16), making it less suitable for minimizing false negatives.
• Key predictors across models included:
  • Revolving Utilization of Unsecured Lines
  • Past Due Counts (30–59, 60–89, 90+)
  • Age

---

🛠️ Tools & Libraries Used

• Python, Jupyter Notebook
• pandas, numpy, seaborn, matplotlib
• scikit-learn, XGBoost
• WOE Binning tools, scikit-plot

---

📁 Repository Structure

credit-risk-prediction/

1. data/ # Raw dataset
2. notebooks/ # Model training and evaluation (LR, RF, XGBoost)
3. results/ # Evaluation metrics, ROC curves, confusion matrices, scorecard
4. README.md # Project overview

---

📌 Conclusion

This project demonstrates that:

• Random Forest is the most effective model for credit delinquency prediction, offering high recall and balanced precision.
• Logistic Regression with WOE remains highly interpretable and practical for deployment via scorecards.
• Combining Random Forest's accuracy with Logistic Regression's explainability provides a strong, business-ready solution for credit scoring systems.