The Topological-Graph Fusion: A Survey of TDA and GNNs for Supply Chain Credit Risk

This survey examines the emerging application of Topological Data Analysis (TDA) and Graph Neural Networks (GNN) in credit risk assessment for Supply Chain Finance (SCF), with particular emphasis on Small and Medium-sized Enterprises (SMEs). Traditional credit risk assessment methods, such as discriminant analysis and logistic regression, often fail to capture the complex, non- linear relationships and systemic interdependencies inherent in modern supply chains. Recent advances in TDA, particularly the BallMapper algorithm, combined with GNN architectures such as GraphSAGE, offer promising solutions by modelling financial indicators as graph structures and capturing multidimensional risk patterns. This survey synthesizes findings from recent studies demonstrating that hybrid TDA-GNN models achieve superior predictive accuracy (92-96%) compared to conventional machine learning approaches, while also providing enhanced interpretability through topological visualization and SHAP analysis. The paper reviews key methodologies, validates their effectiveness across multiple datasets, and discusses practical applications in business loan evaluation, internal management, and investment decision-making. We identify critical research gaps and propose future directions for improving model robustness, scalability, and real-time prediction capabilities in supply chain credit risk management.