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

This paper presented the system design and implementation of a Topological-Graph Fusion model for supply chain credit risk assessment. The implemented BM-GNN architecture achieves the survey's benchmark performance of 93.56% one-year accuracy and 83% two-year accuracy on the COMPUSTAT dataset, representing a significant 11-percentage-point improvement over traditional MDA at the two-year horizon. The key implementation contributions are: (1) a BallMapper pipeline that visualises non-uniform bankruptcy clustering in financial indicator space; (2) a Kendall tau KNN graph that captures implicit firm similarity without transactional data; (3) a three-stage GraphSAGE model that propagates risk signals through supply chain networks; (4) a SHAP interpretability layer meeting regulatory adverse-action requirements; and (5) a three-tier credit decision policy implementing the empirical Z-score cutoff of 2.675. Future work will address: dynamic LSTM-GNN hybrids for multi-year temporal prediction, multi-modal fusion of transaction logs and news sentiment, industry-specific model calibration for technology and service sectors, adversarial robustness via peer consistency checking, and systemic risk modelling for macro-prudential policy support.

[1]     Izhan Shaikh, P.V. Kulkarni, Aman Gokhale, Anchal Jainapure, & Fauzaan Sayyed, “The Topological-Graph Fusion: A Survey of TDA and GNNs for Supply Chain Credit Risk”, International Journal of Scientific Research in Engineering and Management (IJSREM), Volume: 09 Issue: 10 | Oct - 2025, e-ISSN: 2582-3930. [2]     Mojdehi, K.F., Amiri, B., & Haddadi, A. (2025). A Novel Hybrid Model for Credit Risk Assessment of Supply Chain Finance Based on TDA and GNN. IEEE Access, 13, 13101–13127. [3]     Hamilton, W.L., Ying, R., & Leskovec, J. (2017). Inductive Representation Learning on Large Graphs. NeurIPS, 30. [4]     Dlotko, P. (2019). Ball Mapper: A Shape Summary for Topological Data Analysis. arXiv:1901.07410. [5]     Altman, E.I. (1968). Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy. Journal of Finance, 23(4), 589–609. [6]     Ohlson, J.A. (1980). Financial Ratios and the Probabilistic Prediction of Bankruptcy. Journal of Accounting Research, 18(1), 109–131. [7]     Qiu, W., Rudkin, S., & Dlotko, P. (2020). Refining Understanding of Corporate Failure through TDA Mapping of Altman's Z-score Model. Expert Systems with Applications, 156, 113475. [8]     Theunissen, F.M., Alam, S., & Sajjad, A. (2025). An Analytical Framework for Decision Criteria Validation in Complex Supply Chains. Supply Chain Analytics, 12, 100169. [9]     Liu, B., Li, I., Yao, J., et al. (2024). Unveiling the Potential of GNNs in SME Credit Risk Assessment. arXiv:2409.17909. [10] Liu, J. & Kok, S. (2025). Prediction of Bank Credit Ratings using Heterogeneous Topological GNNs. arXiv:2506.06293. [11] Beaver, W.H. (1966). Financial Ratios as Predictors of Failure. Journal of Accounting Research, 4, 71–111. [12] Ghadge, A. et al. (2021). Impact of Financial Risk on Supply Chains: A Manufacturer-Supplier Relational Perspective. International Journal of Production Research, 59(23), 7090–7105. [13] Lundberg, S.M. & Lee, S.I. (2017). A Unified Approach to Interpreting Model Predictions. NeurIPS, 30.