AI2030 Responsible AI Tools Marketplace

Key Takeaways

Privacy protection meets practical utility: The system uses homomorphic encryption within federated learning to allow supply chain partners to collaboratively assess credit risk without sharing sensitive financial data directly—addressing a major barrier to information sharing in competitive supply chain environments.
Explainability is built-in, not added later: Rather than treating interpretability as an afterthought, the architecture incorporates transformer encoders and integrated gradient techniques from the ground up, enabling stakeholders to understand why a credit decision was made rather than just accepting a "black box" recommendation.
Handles real-world data messiness: The performance-based client selection strategy specifically tackles two common problems in federated systems—class imbalance (when defaults are rare) and non-identical data distributions (when different supply chain partners have different customer profiles)—which are critical for reliable credit assessment in practice.
Balances competing demands: The system simultaneously addresses five major challenges (privacy, data silos, class imbalance, data heterogeneity, and interpretability) rather than optimizing for just one, reflecting the complex tradeoffs required in responsible AI deployment for financial decisions.
Validated on real supply chain data: The experimental results demonstrate the approach works on actual datasets rather than just theoretical benchmarks, suggesting practical viability for implementation across supply chain networks.

Abstract

This paper proposes a Trans-XFed architecture that combines federated learning with explainable AI techniques for supply chain credit assessment. The proposed model aims to address several key challenges, including privacy, information silos, class imbalance, non-identically and independently distributed (Non-IID) data, and model interpretability in supply chain credit assessment. We introduce a performance-based client selection strategy (PBCS) to tackle class imbalance and Non-IID problems. This strategy achieves faster convergence by selecting clients with higher local F1 scores. The FedProx architecture, enhanced with homomorphic encryption, is used as the core model, and further incorporates a transformer encoder. The transformer encoder block provides insights into the learned features. Additionally, we employ the integrated gradient explainable AI technique to offer insights into decision-making. We demonstrate the effectiveness of Trans-XFed through experimental evaluations on real-world supply chain datasets. The obtained results show its ability to deliver accurate credit assessments compared to several baselines, while maintaining transparency and privacy.

Accepted by FLTA 2025

Key Takeaways

Privacy protection meets practical utility: The system uses homomorphic encryption within federated learning to allow supply chain partners to collaboratively assess credit risk without sharing sensitive financial data directly—addressing a major barrier to information sharing in competitive supply chain environments.
Explainability is built-in, not added later: Rather than treating interpretability as an afterthought, the architecture incorporates transformer encoders and integrated gradient techniques from the ground up, enabling stakeholders to understand why a credit decision was made rather than just accepting a "black box" recommendation.
Handles real-world data messiness: The performance-based client selection strategy specifically tackles two common problems in federated systems—class imbalance (when defaults are rare) and non-identical data distributions (when different supply chain partners have different customer profiles)—which are critical for reliable credit assessment in practice.
Balances competing demands: The system simultaneously addresses five major challenges (privacy, data silos, class imbalance, data heterogeneity, and interpretability) rather than optimizing for just one, reflecting the complex tradeoffs required in responsible AI deployment for financial decisions.
Validated on real supply chain data: The experimental results demonstrate the approach works on actual datasets rather than just theoretical benchmarks, suggesting practical viability for implementation across supply chain networks.

Abstract

Accepted by FLTA 2025

Trans-XFed: An Explainable Federated Learning for Supply Chain Credit Assessment

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links

Trans-XFed: An Explainable Federated Learning for Supply Chain Credit Assessment

Key Takeaways

Abstract

Paper Info

Responsible AI Topics

Links