Built a Logistic Regression model to predict loan default risk using borrower demographic, financial, and loan-related data.
Dataset obtained from Kaggle: https://www.kaggle.com/datasets/nikhil1e9/loan-default
- Python
- Pandas
- NumPy
- Scikit-learn
- Matplotlib
Created additional risk indicators including:
- LoanToIncomeRatio
- DebtExposure
- InterestBurden
- InterestToIncome
- IncomePerCreditLine
- CreditAdjustedLoan
- ROC-AUC Score: 0.761
- Improved model performance through feature engineering.
- LoanToIncomeRatio
- InterestRate
- NumCreditLines
- EmploymentType_Unemployed
- DTIRatio
- Age
- MonthsEmployed
- HasCoSigner
- HasDependents
- EmploymentType_Full-time
The model successfully identified important drivers of loan default risk and demonstrated good discrimination ability with a ROC-AUC score of 0.761.