Paper: arXiv 2510.09407

Authors: Sahab Zandi, Kamesh Korangi, Juan C. Moreno-Paredes, María Óskarsdóttir, Christophe Mues, Cristián Bravo

Abstract

Small and Medium-sized Enterprises (SMEs) are known to play a vital role in economic growth, employment, and innovation. However, they tend to face significant challenges in accessing credit due to limited financial histories, collateral constraints, and exposure to macroeconomic shocks. These challenges make an accurate credit risk assessment by lenders crucial, particularly since SMEs frequently operate within interconnected firm networks through which default risk can propagate. This paper presents and tests a novel approach for modelling the risk of SME credit, using a unique large data set of SME loans provided by a prominent financial institution. Specifically, our approach employs Graph Neural Networks to predict SME default using multilayer network data derived from common ownership and financial transactions between firms. We show that combining this information with traditional structured data not only improves application scoring performance, but also explicitly models contagion risk between companies. Further analysis shows how the directionality and intensity of these connections influence financial risk contagion, offering a deeper understanding of the underlying processes. Our findings highlight the predictive power of network data, as well as the role of supply chain networks in exposing SMEs to correlated default risk.

Complexity vs Empirical Score

  • Math Complexity: 5.5/10
  • Empirical Rigor: 8.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: This paper presents an innovative multimodal Graph Neural Network framework that incorporates ownership and transaction networks into SME credit risk scoring. It demonstrates strong empirical rigor by utilizing a unique, large-scale proprietary banking dataset to explicitly model financial risk contagion and directionality.

Research Flowchart

  flowchart TD
    A[Research Goal: Improve SME Credit Scoring & Model Contagion Risk] --> B[Data Inputs: Structured Data, Transaction Networks & Ownership Networks]
    B --> C[Methodology: Multimodal Graph Neural Network Model]
    C --> D[Computational Process: Integrate Network Topologies with Financial Features]
    D --> E[Outcome 1: Superior SME Default Prediction Performance]
    D --> F[Outcome 2: Quantification of Risk Contagion via Connection Direction & Intensity]

Cited References (71)

  1. Network science (2016)
  2. Multilayer networks (2014)
  3. Fusion of graph and tabular deep learning models for predicting chronic kidney disease (2023)
  4. The value of text for small business default prediction: A deep learning approach (2021)
  5. A review of affective computing: From unimodal analysis to multimodal fusion (2017)
  6. Credit scoring and its applications (2017)
  7. Evolution of credit risk using a personalized pagerank algorithm for multilayer networks (2020)
  8. Multilayer network analysis for improved credit risk prediction (2021)
and 63 more
  1. Loan default correlation using an A rchimedean copula approach: A case for recalibration (2015)
  2. The value of big data for credit scoring: Enhancing financial inclusion using mobile phone data and social network analytics (2019)
  3. Graph Attention Networks (2018)
  4. How powerful are graph neural networks? (2019)
  5. Financial constraints and small and medium enterprises: A review (2020)
  6. Financial ratios, discriminant analysis and the prediction of corporate bankruptcy (1968)
  7. Clues from networks: Quantifying relational risk for credit risk evaluation of SME s (2022)
  8. On network theory (2011)
  9. A network analysis of the I talian overnight money market (2008)
  10. Cyclical correlations, credit contagion, and portfolio losses (2004)
  11. Group affiliation and default prediction (2019)
  12. The effect of interfirm financial transactions on the credit risk of small and medium-sized enterprises (2019)
  13. Corporate payments networks and credit risk rating (2019)
  14. “ B irds of a feather” fail together: Exploring the nature of dependency in SME defaults (2019)
  15. A literature review of small and medium enterprises ( SME ) risk management practices in S outh A frica (2012)
  16. Deep multimodal learning: A survey on recent advances and trends (2017)
  17. Multimodal machine learning: A survey and taxonomy (2018)
  18. Credit risk prediction of SME s in supply chain finance by fusing demographic and behavioral data (2022)
  19. Multi-modal deep learning for credit rating prediction using text and numerical data streams — → 2304.10740
  20. A transformer-based model for default prediction in mid-cap corporate markets (2023)
  21. Evaluating the credit risk of SME s using legal judgments (2020)
  22. A data-driven and network-aware approach for credit risk prediction in supply chain finance (2021)
  23. A reduction of a graph to a canonical form and an algebra arising during this reduction (1968)
  24. Multi-source heterogeneous data fusion (2018)
  25. Multisensor integration and fusion: Issues and approaches (1988)
  26. Early, intermediate and late fusion strategies for robust deep learning-based multimodal action recognition (2021)
  27. A review on data fusion in multimodal learning analytics and educational data mining (2022)
  28. Multimodal intelligence: Representation learning, information fusion, and applications (2020)
  29. Stacked cross attention for image-text matching (2018)
  30. A unified approach to interpreting model predictions (2017)
  31. Explainable AI for operational research: A defining framework, methods, applications, and a research agenda (2024)
  32. Strictly proper scoring rules, prediction, and estimation (2007)
  33. Evaluating credit risk models (2000)
  34. Financial contagion (2000)
  35. Systemic risk in financial networks: A survey (2021)
  36. Measuring default correlation (2001)
  37. Combining structured and unstructured data for predictive models: A deep learning approach (2020)
  38. Attention-based dynamic multilayer graph neural networks for loan default prediction — → 2402.00299
  39. Access to finance for innovative SME s since the financial crisis (2015)
  40. What are the biggest obstacles to growth of SME s in developing countries? – A n empirical evidence from an enterprise survey (2016)
  41. The effect of financial inclusion on economic growth: The role of human capital development (2024)
  42. Granting and managing loans for micro-entrepreneurs: New developments and practical experiences (2013)
  43. Random forests (2001)
  44. XGB oost: A scalable tree boosting system (2016)
  45. Deep learning (2015)
  46. Graph neural networks based log anomaly detection and explanation (2024) — link
  47. Deep learning for credit scoring: Do or don’t? (2021)
  48. Credit shock propagation along supply chains: Evidence from the CDS market (2022)
  49. Estimating the impact of supply chain network contagion on financial stability — → 2305.04865
  50. Bank default risk propagation along supply chains: Evidence from the UK (2023)
  51. Efficient multimodal learning for corporate credit risk prediction with an extended deep belief network (2025)
  52. Predicting corporate credit risk: Network contagion via trade credit (2021)
  53. Information transfers among co-owned firms (2017)
  54. Identifying discreditable firms in a large-scale ownership network (2022) — link
  55. Time series feature redundancy paradox: An empirical study based on mortgage default prediction — → 2501.00034
  56. Deep neural networks and tabular data: A survey (2022)
  57. Deep multimodal data fusion (2024)
  58. Deep learning models for bankruptcy prediction using textual disclosures (2019)
  59. Applied logistic regression (2013)
  60. The banking transactions dataset and its comparative analysis with scale-free networks (2021)
  61. Anomaly detection in networks with application to financial transaction networks (2019) — link
  62. Customer concentration risk and the cost of equity capital (2016)
  63. Incorporating contagion in portfolio credit risk models using network theory (2018)

4 of 71 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.