A novel method for credit scoring based on feature transformation and ensemble model

Hongxiang Li, Ao Feng, Bin Lin, Houcheng Su, Zixi Liu, Xuliang Duan, Haibo Pu and Yifei Wang

Santa Cruz Department of Economics, Working Paper Series from Department of Economics, UC Santa Cruz

Abstract: Credit scoring is a very critical task for banks and other financial institutions, and it has become an important evaluation metric to distinguish potential defaulting users. In this paper, we propose a credit score prediction method based on feature transformation and ensemble model, which is essentially a cascade approach. The feature transformation process consisting of boosting trees (BT) and auto-encoders (AE) is employed to replace manual feature engineering and to solve the data imbalance problem. For the classification process, this paper designs a heterogeneous ensemble model by weighting the factorization machine (FM) and deep neural networks (DNN), which can efficiently extract low-order intersections and high-order intersections. Comprehensive experiments were conducted on two standard datasets and the results demonstrate that the proposed approach outperforms existing credit scoring models in accuracy.

Keywords: AutoEncoder; Boosting tree; Credit scoring; Deep neural network; Factorization machine; Feature transformation (search for similar items in EconPapers)
Date: 2021-01-01
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
https://www.escholarship.org/uc/item/3v33k65c.pdf;origin=repeccitec (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:cdl:ucscec:qt3v33k65c

Access Statistics for this paper

More papers in Santa Cruz Department of Economics, Working Paper Series from Department of Economics, UC Santa Cruz Contact information at EDIRC.
Bibliographic data for series maintained by Lisa Schiff ().