Loss given default estimation: A two-stage model with classification tree-based boosting and support vector logistic regression

Yuta Tanoue, Satoshi Yamashita

研究成果: Article

1 引用 (Scopus)

抄録

The Basel Accords allow banks to estimate credit risk. Accordingly, more attention has been dedicated recently to the analysis of loss given default (LGD) and the development of an LGD estimation model. In this study, using a data set composed of five Japanese regional banks, we propose an LGD estimation model using a twostage model, classification tree-based boosting and support vector regression (SVR). We compare the proposed model’s predictive performance with existing models by performing cross-validation and out-of-time validation. As a result, we find that incorporating nonlinearity into the LGD estimation model by classification and SVR improves its predictive performance. Further, we confirm that the boosting method improves the model’s predictive performance.

元の言語English
ページ(範囲)19-37
ページ数19
ジャーナルJournal of Risk
21
発行部数4
DOI
出版物ステータスPublished - 2019 4
外部発表Yes

Fingerprint

Loss given default
Two-stage model
Logistic regression
Boosting
Support vector regression
Basel Accord
Nonlinearity
Credit risk
Cross-validation

ASJC Scopus subject areas

  • Finance
  • Strategy and Management

これを引用

@article{f35be661a8344a81971aa0bcfedfa030,
title = "Loss given default estimation: A two-stage model with classification tree-based boosting and support vector logistic regression",
abstract = "The Basel Accords allow banks to estimate credit risk. Accordingly, more attention has been dedicated recently to the analysis of loss given default (LGD) and the development of an LGD estimation model. In this study, using a data set composed of five Japanese regional banks, we propose an LGD estimation model using a twostage model, classification tree-based boosting and support vector regression (SVR). We compare the proposed model’s predictive performance with existing models by performing cross-validation and out-of-time validation. As a result, we find that incorporating nonlinearity into the LGD estimation model by classification and SVR improves its predictive performance. Further, we confirm that the boosting method improves the model’s predictive performance.",
keywords = "Boosting, Credit risk, Loss given default (LGD), Platt scaling, Risk management",
author = "Yuta Tanoue and Satoshi Yamashita",
year = "2019",
month = "4",
doi = "10.21314/JOR.2019.405",
language = "English",
volume = "21",
pages = "19--37",
journal = "Journal of Risk",
issn = "1465-1211",
publisher = "Incisive Media Ltd.",
number = "4",

}

TY - JOUR

T1 - Loss given default estimation

T2 - A two-stage model with classification tree-based boosting and support vector logistic regression

AU - Tanoue, Yuta

AU - Yamashita, Satoshi

PY - 2019/4

Y1 - 2019/4

N2 - The Basel Accords allow banks to estimate credit risk. Accordingly, more attention has been dedicated recently to the analysis of loss given default (LGD) and the development of an LGD estimation model. In this study, using a data set composed of five Japanese regional banks, we propose an LGD estimation model using a twostage model, classification tree-based boosting and support vector regression (SVR). We compare the proposed model’s predictive performance with existing models by performing cross-validation and out-of-time validation. As a result, we find that incorporating nonlinearity into the LGD estimation model by classification and SVR improves its predictive performance. Further, we confirm that the boosting method improves the model’s predictive performance.

AB - The Basel Accords allow banks to estimate credit risk. Accordingly, more attention has been dedicated recently to the analysis of loss given default (LGD) and the development of an LGD estimation model. In this study, using a data set composed of five Japanese regional banks, we propose an LGD estimation model using a twostage model, classification tree-based boosting and support vector regression (SVR). We compare the proposed model’s predictive performance with existing models by performing cross-validation and out-of-time validation. As a result, we find that incorporating nonlinearity into the LGD estimation model by classification and SVR improves its predictive performance. Further, we confirm that the boosting method improves the model’s predictive performance.

KW - Boosting

KW - Credit risk

KW - Loss given default (LGD)

KW - Platt scaling

KW - Risk management

UR - http://www.scopus.com/inward/record.url?scp=85067041651&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85067041651&partnerID=8YFLogxK

U2 - 10.21314/JOR.2019.405

DO - 10.21314/JOR.2019.405

M3 - Article

AN - SCOPUS:85067041651

VL - 21

SP - 19

EP - 37

JO - Journal of Risk

JF - Journal of Risk

SN - 1465-1211

IS - 4

ER -