Credit Risk Modelling Project: PD, LGD, EAD This repository contains a comprehensive end-to-end implementation of a Credit Risk Modelling framework, focusing on the development of the three key risk components used under Basel II/III Internal Ratings-Based (IRB) approaches:

Probability of Default (PD)

Loss Given Default (LGD)

Exposure at Default (EAD)

The models are developed using Python, based on real-world-inspired datasets (e.g., Lending Club) and follow a practical industry-standard methodology, useful for credit analysts, data scientists, and risk modelers.

Project Objectives

Understand and model the credit risk of individual loans

Estimate the likelihood of default (PD), expected losses (LGD), and outstanding exposure (EAD)

Apply industry practices such as:

Weight of Evidence (WoE) & Information Value (IV)

Logistic and Linear Regression modeling

Feature binning & risk grouping

Model evaluation (AUC, Gini, KS, RMSE, R²)

Regulatory-compliant model structure

Project Structure Section Description Data Preprocessing Handling missing values, outliers, variable transformations PD Modeling Fine & coarse classing, WoE encoding, logistic regression, evaluation LGD Modeling Regression on recovery rate, linear regression, log transformations EAD Modeling Usage given default modeling using linear regression Model Validation KS-statistic, AUC, ROC curves, residual analysis

Technologies Used Python: Data preprocessing, modeling, and evaluation

Libraries: pandas, numpy, matplotlib, seaborn, sklearn, statsmodels

Key Techniques Weight of Evidence (WoE) Transformation: Converts categorical variables for PD modeling

Information Value (IV): Feature selection based on predictive power

Binning Techniques: Fine and coarse classing for continuous variables

Model Evaluation: ROC, AUC, KS, adjusted R², RMSE

Insights Gained PD model helps estimate the probability of default using borrower attributes.

LGD model provides insight into how much is lost if a default occurs.

EAD model helps predict the exposure amount at the point of default.

Combining these three gives the Expected Loss (EL): EL = PD × LGD × EAD

How to Use Clone this repo:

bash Copy Edit git clone https://github.com/yourusername/credit-risk-modeling.git Install required dependencies:

bash Copy Edit pip install -r requirements.txt Run the Jupyter Notebook:

bash Copy Edit jupyter notebook Credit\ Risk\ Modelling\ Project\ PD\ LGD\ EAD.ipynb License This project is open-source and available under the MIT License.

Acknowledgements Inspired by industry practices and academic methodologies

Datasets inspired by platforms like Lending Club and public data repositories