Economic insights into credit risk: a deep learning model for predicting loan repayment capacity

Accurately predicting loan repayment capacity is critical to effective credit risk assessment in banking, finance, and economic policymaking. This study proposes a novel deep learning architecture, the Gated Transformer, designed specifically for structured credit data. The model integrates categorical and numerical features into a unified latent space and employs multi-head self-attention combined with a token-wise gating mechanism to dynamically capture complex feature interactions while enhancing interpretability. Experiments conducted on the LendingClub dataset demonstrate that the proposed model achieves strong predictive performance, with an accuracy of 92.38%, an area under the receiver operating characteristic curve (AUROC) of 0.8517, and an area under the precision-recall curve (AUPRC) of 0.9213. Further robustness tests across multiple random seeds confirm the model’s stability. These findings underscore the model’s effectiveness in handling heterogeneous financial data and its potential to support responsible, data-driven credit risk evaluation.