Credit Risk Prediction

Project Overview

Machine learning project to predict loan default risk using the loan_data_2007_2014 dataset.
Covers data cleaning, feature engineering, EDA, preparation, model training, and evaluation.
https://drive.google.com/drive/folders/1demEFp_i47EC8sJCocU9ubAeEQthUYR_?usp=drive_link

---

Problem Statement

Financial institutions need to assess loan default risk before approval.
Goal: classify loans into:

• Fully Paid (0)
• Charged Off (1)

---

Methods

• Models: Logistic Regression (baseline) & Random Forest.
• Evaluation Metrics: Accuracy, Precision, Recall, F1-score, ROC-AUC.

---

Results

• Logistic Regression → Accuracy: ~100%, ROC-AUC: 0.999
• Random Forest → Accuracy: ~100%, ROC-AUC: 1.000
• Random Forest chosen as final model due to slightly better performance.

---

Conclusion

The project shows that interest rate, grade, and DTI are key indicators of loan default risk.
Random Forest provides the most consistent results and can support better credit risk management.