Bridging accuracy and interpretability: A rescaled cluster-then-predict approach for enhanced credit scoring

Huei-Wen Teng, Ming-Hsuan Kang, I-Han Lee and Le-Chi Bai

International Review of Financial Analysis, 2024, vol. 91, issue C

Abstract: Credit scoring is pivotal in the financial industry for assessing individuals’ creditworthiness and optimizing financial institutions’ risk-adjusted returns. While the XGBoost algorithm stands as the state-of-the-art classifier for credit scoring, its intricate nature impedes easy interpretation, a critical aspect for stakeholders’ decision-making. This paper introduces a novel approach termed the “Rescaled Cluster-then-Predict Method,” aimed at enhancing both the interpretability and predictive performance of credit scoring models. Our method employs a two-step process, initially rescaling the features and subsequently clustering the data into subgroups. Consequently, we employ Logistic Regression within each subgroup to generate predictions. The paper’s primary contributions are twofold. Firstly, empirical evaluations on two distinct datasets demonstrate that our proposed method attains a competitive performance compared to XGBoost while substantially improving interpretability. Notably, in some instances, the Logistic Regression outperforms XGBoost. Secondly, we reveal that clustering solely the positive cases, as opposed to the entire dataset, yields comparable results while markedly reducing computational requirements. These insights hold significant practical implications for the financial industry, which consistently seeks credit scoring models that are not only accurate but also interpretable and computationally efficient.

Keywords: Credit scoring; Cluster-then-predict; Rescaling; XGBoost; Logistic Regression (search for similar items in EconPapers)
JEL-codes: C38 C53 G21 (search for similar items in EconPapers)
Date: 2024
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S1057521923005215
Full text for ScienceDirect subscribers only

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:eee:finana:v:91:y:2024:i:c:s1057521923005215

DOI: 10.1016/j.irfa.2023.103005

Access Statistics for this article

International Review of Financial Analysis is currently edited by B.M. Lucey

More articles in International Review of Financial Analysis from Elsevier
Bibliographic data for series maintained by Catherine Liu ().