A Multimodal Approach to SME Credit Scoring Integrating Transaction and Ownership Networks

TL;DR

The paper addresses SME credit risk by introducing a multimodal framework that integrates multilayer networks derived from financial transactions and ownership with traditional borrower data. Using Graph Attention Networks and Graph Isomorphism Networks, coupled with early/intermediate fusion and cross-attention, the approach learns network-aware embeddings and yields superior default predictions compared to baselines. Key findings show that bimodal models, especially with a hybrid fusion strategy and directed/weighted edges, substantially improve AUC and AUCPR while providing interpretable insights into contagion mechanisms via SHAP analyses. This work demonstrates the practical value of explicitly modelling inter-firm contagion in credit risk and lays groundwork for broader adoption across lending contexts and additional data modalities.

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.

Figures (11)

• Figure 1: A multilayer network (left) and its supra adjacency matrix (right). Available under the CC-BY license from zandi2024attention.
• Figure 2: The three primary fusion levels: early, intermediate, and late.
• Figure 3: The proposed fusion strategies for the multimodal models. Adapted from tavakoli2023multi.
• Figure 4: An overview of the architectures of networks A and B.
• Figure 5: The unimodal and bimodal model structures.