LOTUS-based segmentation in credit scoring
LOTUS-based segmentation in credit scoring
Credit scoring allows for the credit risk assessment of bank customers. A single scoring model (scorecard) can be built for the entire customer population. However, dividing the population into several groups and building separate scorecards for them can yield better results (in particular, higher performance of the whole model). That division is referred to as segmentation and is widely used in banking practice. There are various segmentation methods. A few methods have recently been developed in an attempt to enable the optimal segmentation, i.e. such segmentation that would maximise the model performance. However, those methods still suffer from serious drawbacks such as the exhaustive search, predetermined number of segments or using the same set of variables in all scorecards. In this research, a new segmentation method is suggested, the Logistic Tree with Unbiased Selection (LOTUS) algorithm. The suggested method is derived from data mining. It is free from the above-mentioned drawbacks and allows for the optimal segmentation. It is tested using the data provided by one of the major UK banks and one of the European credit bureaus. For comparison purposes, some reference models are also developed using techniques that are popular in credit scoring (logistic regression and classification trees)
Bijak, Katarzyna
5130b6b9-fbf1-44e8-9106-1dd69c6692a6
2010
Bijak, Katarzyna
5130b6b9-fbf1-44e8-9106-1dd69c6692a6
Bijak, Katarzyna
(2010)
LOTUS-based segmentation in credit scoring.
24th European Conference on Operational Research (EURO XXIV), , Lisbon, Portugal.
10 - 13 Jul 2010.
Record type:
Conference or Workshop Item
(Other)
Abstract
Credit scoring allows for the credit risk assessment of bank customers. A single scoring model (scorecard) can be built for the entire customer population. However, dividing the population into several groups and building separate scorecards for them can yield better results (in particular, higher performance of the whole model). That division is referred to as segmentation and is widely used in banking practice. There are various segmentation methods. A few methods have recently been developed in an attempt to enable the optimal segmentation, i.e. such segmentation that would maximise the model performance. However, those methods still suffer from serious drawbacks such as the exhaustive search, predetermined number of segments or using the same set of variables in all scorecards. In this research, a new segmentation method is suggested, the Logistic Tree with Unbiased Selection (LOTUS) algorithm. The suggested method is derived from data mining. It is free from the above-mentioned drawbacks and allows for the optimal segmentation. It is tested using the data provided by one of the major UK banks and one of the European credit bureaus. For comparison purposes, some reference models are also developed using techniques that are popular in credit scoring (logistic regression and classification trees)
This record has no associated files available for download.
More information
Published date: 2010
Venue - Dates:
24th European Conference on Operational Research (EURO XXIV), , Lisbon, Portugal, 2010-07-10 - 2010-07-13
Organisations:
Centre of Excellence for International Banking, Finance & Accounting
Identifiers
Local EPrints ID: 361327
URI: http://eprints.soton.ac.uk/id/eprint/361327
PURE UUID: 2efeded5-5f11-474e-926c-9658dead4cbb
Catalogue record
Date deposited: 23 Jan 2014 10:21
Last modified: 11 Dec 2021 04:27
Export record
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics
