Use this URL to cite or link to this record in EThOS:
Title: Basel II compliant credit risk modelling : model development for imbalanced credit scoring data sets, loss given default (LGD) and exposure at default (EAD)
Author: Brown, Iain Leonard Johnston
ISNI:       0000 0004 2726 7127
Awarding Body: University of Southampton
Current Institution: University of Southampton
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
Access from Institution:
The purpose of this thesis is to determine and to better inform industry practitioners to the most appropriate classification and regression techniques for modelling the three key credit risk components of the Basel II minimum capital requirement; probability of default (PD), loss given default (LGD), and exposure at default (EAD). The Basel II accord regulates risk and capital management requirements to ensure that a bank holds enough capital proportional to the exposed risk of its lending practices. Under the advanced internal ratings based (IRB) approach Basel II allows banks to develop their own empirical models based on historical data for each of PD, LGD and EAD. In this thesis, first the issue of imbalanced credit scoring data sets, a special case of PD modelling where the number of defaulting observations in a data set is much lower than the number of observations that do not default, is identified, and the suitability of various classification techniques are analysed and presented. As well as using traditional classification techniques this thesis also explores the suitability of gradient boosting, least square support vector machines and random forests as a form of classification. The second part of this thesis focuses on the prediction of LGD, which measures the economic loss, expressed as a percentage of the exposure, in case of default. In this thesis, various state-of-the-art regression techniques to model LGD are considered. In the final part of this thesis we investigate models for predicting the exposure at default (EAD). For off-balance-sheet items (for example credit cards) to calculate the EAD one requires the committed but unused loan amount times a credit conversion factor (CCF). Ordinary least squares (OLS), logistic and cumulative logistic regression models are analysed, as well as an OLS with Beta transformation model, with the main aim of finding the most robust and comprehensible model for the prediction of the CCF. Also a direct estimation of EAD, using an OLS model, will be analysed. All the models built and presented in this thesis have been applied to real-life data sets from major global banking institutions.
Supervisor: Mues, Christophe Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: HF Commerce ; HG Finance