Post-processing based neural networks for credit scoring

Yan Jiao; Kin Keung Lai; Xia Ge; Liang Liang

Post-processing based neural networks for credit scoring

Yan Jiao, Kin Keung Lai, Xia Ge, Liang Liang

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

Abstract

Due to the existence of failed trained samples, the possibility of samples being wrongly classified, for prediction, increases because of the similarity between forecasting samples and the training samples. To avoid this potential error, in this study, a two-stage hybrid model, which introduces a post-processing distinguish process after the ANN training, is proposed to identify the failed trained samples and filter out unsafe forecasting samples. This provides higher prediction accuracy. At last a real-world dataset is used to prove that this is a workable alternative model for credit scoring tasks.

Original language	English
Title of host publication	Advances in Applied Computing and Computational Sciences - Proceedings of International Symposium on Applied Computing and Computational Sciences, ACCS 2008
Publisher	Global Information Publisher (H.K) Co., Limited
Pages	138-143
ISBN (Print)	9789889964405
Publication status	Published - 2008
Event	2008 International Symposium on Applied Computing and Computational Sciences, ACCS 2008 - Hong Kong, China Duration: 1 Aug 2008 → 3 Aug 2008

Conference

Conference	2008 International Symposium on Applied Computing and Computational Sciences, ACCS 2008
Place	China
City	Hong Kong
Period	1/08/08 → 3/08/08

Research Keywords

Classification Score
Distinguish Process
Neural Network
Similarity

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@inproceedings{bafb5f23825144c68ce7a76404035b49,

title = "Post-processing based neural networks for credit scoring",

abstract = "Due to the existence of failed trained samples, the possibility of samples being wrongly classified, for prediction, increases because of the similarity between forecasting samples and the training samples. To avoid this potential error, in this study, a two-stage hybrid model, which introduces a post-processing distinguish process after the ANN training, is proposed to identify the failed trained samples and filter out unsafe forecasting samples. This provides higher prediction accuracy. At last a real-world dataset is used to prove that this is a workable alternative model for credit scoring tasks.",

keywords = "Classification Score, Distinguish Process, Neural Network, Similarity",

author = "Yan Jiao and Lai, {Kin Keung} and Xia Ge and Liang Liang",

year = "2008",

language = "English",

isbn = "9789889964405",

pages = "138--143",

booktitle = "Advances in Applied Computing and Computational Sciences - Proceedings of International Symposium on Applied Computing and Computational Sciences, ACCS 2008",

publisher = "Global Information Publisher (H.K) Co., Limited",

note = "2008 International Symposium on Applied Computing and Computational Sciences, ACCS 2008 ; Conference date: 01-08-2008 Through 03-08-2008",

}

Jiao, Y, Lai, KK, Ge, X & Liang, L 2008, Post-processing based neural networks for credit scoring. in Advances in Applied Computing and Computational Sciences - Proceedings of International Symposium on Applied Computing and Computational Sciences, ACCS 2008. Global Information Publisher (H.K) Co., Limited, pp. 138-143, 2008 International Symposium on Applied Computing and Computational Sciences, ACCS 2008, Hong Kong, China, 1/08/08.

Post-processing based neural networks for credit scoring. / Jiao, Yan; Lai, Kin Keung; Ge, Xia et al.
Advances in Applied Computing and Computational Sciences - Proceedings of International Symposium on Applied Computing and Computational Sciences, ACCS 2008. Global Information Publisher (H.K) Co., Limited, 2008. p. 138-143.

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - Post-processing based neural networks for credit scoring

AU - Jiao, Yan

AU - Lai, Kin Keung

AU - Ge, Xia

AU - Liang, Liang

PY - 2008

Y1 - 2008

N2 - Due to the existence of failed trained samples, the possibility of samples being wrongly classified, for prediction, increases because of the similarity between forecasting samples and the training samples. To avoid this potential error, in this study, a two-stage hybrid model, which introduces a post-processing distinguish process after the ANN training, is proposed to identify the failed trained samples and filter out unsafe forecasting samples. This provides higher prediction accuracy. At last a real-world dataset is used to prove that this is a workable alternative model for credit scoring tasks.

AB - Due to the existence of failed trained samples, the possibility of samples being wrongly classified, for prediction, increases because of the similarity between forecasting samples and the training samples. To avoid this potential error, in this study, a two-stage hybrid model, which introduces a post-processing distinguish process after the ANN training, is proposed to identify the failed trained samples and filter out unsafe forecasting samples. This provides higher prediction accuracy. At last a real-world dataset is used to prove that this is a workable alternative model for credit scoring tasks.

KW - Classification Score

KW - Distinguish Process

KW - Neural Network

KW - Similarity

UR - http://www.scopus.com/inward/record.url?scp=84945904640&partnerID=8YFLogxK

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-84945904640&origin=recordpage

M3 - RGC 32 - Refereed conference paper (with host publication)

SN - 9789889964405

SP - 138

EP - 143

BT - Advances in Applied Computing and Computational Sciences - Proceedings of International Symposium on Applied Computing and Computational Sciences, ACCS 2008

PB - Global Information Publisher (H.K) Co., Limited

T2 - 2008 International Symposium on Applied Computing and Computational Sciences, ACCS 2008

Y2 - 1 August 2008 through 3 August 2008

ER -

Post-processing based neural networks for credit scoring

Abstract

Conference

Research Keywords

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this