How can consumers without credit history benefit from the use of information processing and machine learning tools by financial institutions - credit risk assessment study

Aug 6, 2026

Upload by Eliana

This research enhances predictability of creditworthiness for marginalized consumers who lack credit history amid widespread AI adoption. Ensemble methods are applied to manage imbalanced data for credit risk evaluation when histories are sparse or absent. To support fairness, the study uses disparate impact remover to reduce group bias in the machine learning model. Dataset imbalance is addressed via oversampling, undersampling, and class-weight adjustment, with class-weight tuning achieving the strongest performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages22

Reading Time55 min

AccessFree to download

Introduction

Scope of the study

Credit risk in short-term loans

Machine learning models and gaps

Methods

Ensemble learning on imbalanced data

Bias mitigation with disparate impact remover

Imbalance handling strategies

Results and implications

Performance of class-weight adjustment

Trade-offs between fairness and predictive power

Information Processing and Management 62 (2025) 103972 | How can consumers without credit history benefit from the use of information processing and machine learning tools by financial institutions? | | | | | --- | --- | --- | --- | | Bjorn van Braak, Joerg R. Osterrieder, Marcos R. Machado ∗
Department of Industrial Engineering and Business Information Systems, University of Twente, 7500 AE Enschede, The Netherlands | | | | | A R T I C L E I N F O | | A B S T R A C T | | | Keywords:
Consumers’ credit risk assessment Information processing Machine learning
Ensemble learning
Financial inclusion | | This research aims to enhance the predictability of creditworthiness among marginalized consumers affected by the widespread adoption of AI frameworks. We utilize ensemble methods to handle the imbalanced dataset used for evaluating the credit risk of consumers with sparse or non-existent credit histories. To promote fairness in the Machine Learning (ML) model, we employed the disparate impact remover—a recognized bias mitigation tool to minimize group bias. Three strategies were employed to tackle dataset imbalance: oversampling, undersampling, and class weight adjustment. Our findings reveal that adjusting the class weight proved most effective in sustaining commendable performance, demonstrating higher accuracy and F-1 scores surpassing 80% in most experiments. While the application of the disparate impact remover might compromise the ML model’s predictive capabilities, our results underscore the necessity of deliberating over the use of potentially bias-sensitive, unprotected features. Recognizing the critical nature of this trade-off for financial decision-makers, we delve into its implications. | | 1. Introduction The ever-growing abundance of data, combined with advancements in computational power and innovative techniques, has transformed approaches to credit risk determination. Over time, the shift from traditional statistical models to Machine Learning (ML) frameworks has become evident (Galindo & Tamayo, 2000). Credit risk research encompasses diverse risks faced by lenders, including credit risk assessment, non-performing asset forecasting, and fraud detection (Bhatore, Mohan, & Reddy, 2020). This study, however, will exclusively focus on the evaluation of credit risk. Various methods exist to assess this risk, such as credit score assignment, gauging the likelihood of loan default, or categorizing lenders into distinct groups based on credit reliability (Kritzinger & Van Vuuren, 2018). Present market models assessing credit risk employ both ML and traditional statistical frameworks. Primarily, these models are grounded in credit-specific information, e.g., credit history, outstanding credit, and the nature of the debt incurred. Our study aims to evaluate the risk associated with short-term loans for retail borrowers who possess limited credit history—a scenario often observed in developing countries or among lower-income clienteles (The World Bank, 2022). Customers in this context frequently lack a substantive financial track record, thereby presenting a challenge for lenders in evaluating their creditworthiness. This heightened credit risk bears critical implications for several reasons (Allen, DeLong, & Saunders, 2004). Firstly, short-term loans inherently encompass compressed repayment periods, thereby elevating the urgency of repayment. Consequently, borrowers with limited ∗ Corresponding author. E-mail addresses: [b.vanbraak@student.utwente.nl](b.vanbraak@student.utwente.nl) (B. van Braak), [joerg.osterrieder@utwente.nl](joerg.osterrieder@utwente.nl) (J.R. Osterrieder), [m.r.machado@utwente.nl](m.r.machado@utwente.nl)[ ](m.r.machado@utwente.nl)(M.R. Machado). [https://doi.org/10.1016/j.ipm.2024.103972](https://doi.org/10.1016/j.ipm.2024.103972) Received 22 October 2023; Received in revised form 1 July 2024; Accepted 8 November 2024 Available online 22 November 2024 0306-4573/©

FAQs

How does the study evaluate credit risk for consumers without credit history?

It focuses on credit risk assessment for retail borrowers with limited or nonexistent credit histories and uses ensemble learning methods to model their creditworthiness.

What approach is used to address dataset imbalance in the machine learning model?

The study applies oversampling, undersampling, and class weight adjustment to tackle imbalanced targets during training and evaluation.

How is fairness handled in the machine learning framework?

Fairness is promoted using the disparate impact remover, a bias-mitigation tool aimed at minimizing group bias.

Aug 6, 2026

Upload by Eliana