Credit Risk Modeling with Python (PD, LGD, EAD)

Credit Risk Modeling using Machine Learning with Python

This project explores a large consumer loan dataset from Lending Club and applies a full risk modeling pipeline using supervised machine learning techniques.

Models Built:

• PD (Probability of Default) – using logistic regression
• LGD (Loss Given Default) – two-stage model
• EAD (Exposure at Default) – linear regression

A FICO-style scorecard was also developed to summarize risk scores for each loan.

📊 Key Features:

• Over 800,000 loans cleaned and processed
• Missing value treatment using correlated features
• Log transformation for skewed variables
• PSI analysis to monitor data drift
• Expected Loss (EL) calculated using PD × LGD × EAD

Tools Used:

• Python (pandas, scikit-learn, matplotlib, plotly)
• Jupyter Notebooks for analysis
• Flask (optional) for model deployment

Business Value:

• Help lenders evaluate borrower risk
• Automate and monitor model stability
• Provide a scorecard interpretable by business teams

Notebooks Included:

• General preprocessing
• PD model data prep + training
• Scorecard development
• PSI monitoring
• LGD & EAD modeling with expected loss

View notebook on GitHub:
🔗 Expected Loss Estimation and Credit Risk Analysis

Full project repo:
🔗 GitHub Repository

Data + models:
🔗 Google Drive Folder