Hybrid boosted attention-based LightGBM framework for enhanced credit risk assessment in digital finance

Chengwei Ying, Anlu Shi and Xiongyi Li ()
Additional contact information
Chengwei Ying: Zhejiang Financial College
Anlu Shi: CIC Japan Innovation Services, University of Massachusetts Lowell
Xiongyi Li: China Executive Leadership Academy Pudong

Palgrave Communications, 2025, vol. 12, issue 1, 1-13

Abstract: Abstract In the rapidly evolving landscape of digital finance, accurate credit risk assessment is critical for mitigating financial risks and improving lending decision-making. Traditional credit scoring methods often struggle with high-dimensional data, class imbalance, and limited interpretability, reducing their effectiveness in predicting borrower default risks. To address these challenges, this study proposes an enhanced Hybrid Boosted Attention-based LightGBM (HBA-LGBM) framework. The HBA-LGBM model introduces four key innovations: (1) a multi-stage feature selection mechanism that dynamically filters key borrower attributes; (2) an attention-based feature enhancement layer, which prioritizes critical financial risk factors dynamically based on contextual importance; (3) a hybrid boosting strategy, integrating LightGBM with an adaptive neural network, enabling the model to capture complex borrower behavior and non-linear credit risk patterns; and (4) an advanced imbalanced learning strategy, combining synthetic data augmentation and cost-sensitive learning to mitigate class imbalance and enhance minority class predictions. To evaluate the effectiveness of HBA-LGBM, experiments were conducted using a large-scale LendingClub online loan dataset. The model was compared with five state-of-the-art methods. The results demonstrate that HBA-LGBM achieves the lowest RMSE (11.53) and MAPE (4.44%), with an R2 score of 0.998, outperforming deep learning and ensemble-based approaches. The model’s superior performance is attributed to its ability to adaptively refine borrower risk assessment, effectively balance computational efficiency with model interpretability, and provide a robust, scalable solution for digital finance applications. This research contributes to the advancement of hybrid machine learning techniques in financial risk management, offering an effective and interpretable approach to credit risk evaluation in digital lending platforms.

Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1057/s41599-025-05230-y Abstract (text/html)
Access to full text is restricted to subscribers.

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:pal:palcom:v:12:y:2025:i:1:d:10.1057_s41599-025-05230-y

Ordering information: This journal article can be ordered from
https://www.nature.com/palcomms/about

DOI: 10.1057/s41599-025-05230-y

Access Statistics for this article

More articles in Palgrave Communications from Palgrave Macmillan
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().