Title: A comparative study of state of the art classification algorithms for credit scoring
Authors: Baesens, Bart
Viaene, Stijn
Vanthienen, Jan #
Issue Date: 2001
Host Document: Proceedings of the Seventh Conference on Credit Scoring and Credit Control (CSCCVII)
Conference: Conference on Credit Scoring and Credit Control (CSCCVII) edition:7 location:Edinburgh (Scotland) date:September 2001
Abstract: In this paper, we study the performance of various state of the art classification algorithms applied to four real-life credit scoring data sets. Two data sets originate from major Benelux financial institutions. Different types of classifiers are evaluated and contrasted. Besides the well-known classification algorithms (e.g. logistic regression, discriminant analysis, k nearest-neighbour, neural networks and decision trees), this study also investigates the suitability and performance of some recently proposed, advanced pattern recognition algorithms (e.g. support vector machines and Bayesian network classifiers) for credit scoring. Experiments are conducted using both the continuous and discretised versions of all data sets. Performance is quantified using both classification accuracy and the area under the receiver operating characteristic curve (AUROC). Statistically significant performance differences are identified using the appropriate test statistics. For the continuous data sets, it was found that the linear discriminant, logistic regression, neural network and support vector machine classifier achieved the best performance in terms of PCC and AUROC on all four data sets. The support vector machine classifier was ranked first on all data sets for both PCC and AUROC. For the discretised data sets, it was found that the linear discriminant and support vector machine classifier gave good performance in terms of both PCC and AUROC while the performance of some of the other classifiers strongly depended upon the performance criterion used.
Publication status: published
KU Leuven publication type: IMa
Appears in Collections:Research Center for Management Informatics (LIRIS), Leuven
# (joint) last author

Files in This Item:

There are no files associated with this item.


All items in Lirias are protected by copyright, with all rights reserved.