AI's Revolution in Credit Risk: From Traditional Models to Neural Networks

Artificial Intelligence (AI) is transforming credit risk modeling by offering methods and tools that serve as a complement - or even an alternative - to traditional statistical approaches such as logistic regression (see Shi et al., 2022 for a systematic review; Bussman et al., 2021). The integration of machine learning in credit risk analysis provides enhanced predictive power, especially in capturing nonlinear relationships and complex interactions among financial variables (Hajj & Hammoud, 2023). Recent literature argues that neural networks and other AI techniques outperform conventional methods in areas such as default prediction, credit scoring, and early warning systems (Byanjankar et al., 2015; Li et al., 2016; Nguyen et al., 2023). However, challenges remain regarding interpretability, regulatory compliance, and data quality, especially when such methods are implemented by supervised financial intermediaries (The Royal Society, 2019; Hadji-Misheva and Osterrieder, 2023). Recent research emphasizes the trade-off between model complexity and transparency, prompting the development of explainable AI techniques to meet financial regulatory standards (Bussman et al., 2025; Fritz-Morgenthal et al., 2022).

This Research Topic searches for high-quality contributions that develop and/or apply AI methods in credit risk assessment, modelling, and management. Indeed, while traditional credit rating approaches remain foundational, the growing complexity and interconnectivity of financial systems call for more adaptive and accurate tools. We also welcome research that goes beyond borrower-level credit scoring to address network effects, contagion channels, and systemic risk, using AI methodologies to investigate financial interdependencies and predict stress propagation across institutions and markets.

As machine learning methods grow in complexity, so does the potential for model instability, overfitting, and lack of transparency. Thus, another important area of interest is the assessment and management of model risk arising from the use of AI techniques in credit risk applications. We also encourage submissions that critically compare traditional statistical approaches to credit risk (such as logistic regression or discriminant analysis) with AI-driven methods, not only in terms of predictive performance, but also of interpretability. With respect to the latter, and given the growing emphasis on trustworthy and responsible AI in financial decision-making, a further objective is indeed to promote the development of transparent and explainable AI (XAI) methods in credit risk analysis.

Finally, we also encourage submissions that investigate the application of AI models in emerging economies and compare the state of the art of AI applications in the field of lending within the traditional channels (e.g., banking channels) and alternative channels introduced with Fintech development (e.g., peer-to-peer lending platforms).

Topics may include, but are not limited to, the following:

- AI-based credit scoring models;
- Modeling systemic risk and network effects using AI techniques;
- Credit risk scenario analysis, stress testing and forecasting using AI;
- Comparative studies between AI methods and traditional credit risk models;
- Fairness, interpretability and explainability of AI models in credit risk;
- Model risk management for AI in credit risk: validation, monitoring, and uncertainty quantification;
- New regulatory challenges in banking credit risk assessment;
- Case studies of real-world AI applications in banking and fintech;
- Hybrid models combining AI with statistical or econometric approaches;
- Data quality, feature selection and dimensionality reduction in credit risk modeling;
- Application of AI in emerging market environments;
- Drivers of AI adoption in emerging markets and developed countries.

We welcome both theoretical and empirical studies that contribute to advancing the understanding and practical implementation of AI in the credit risk management context.