A Blockchain and Federated Learning Based Model for Supply Chain Credit Risk Assessment

This paper proposes a novel blockchain and federated learning fusion model for supply chain credit risk assessment. The focus is on combining privacy protection with collaborative learning. The model combines federated learning with blockchain technology. Multiple participants train a global model collaboratively without sharing raw data. Blockchain technology ensures data im-mutability and transparency. This fusion protects data privacy and ensures the integrity of the collaborative training process. The study evaluates the model's performance in terms of accuracy, privacy protection, and system performance. It also compares the model with traditional centralized and federated learning models. Experimental results show that the proposed model has significant advantages in privacy protection. The use of differential privacy and blockchain immutability effectively reduces the risk of data leakage. However, there is a tradeoff between privacy protection and model performance. The integration of blockchain slightly affects model accuracy. Furthermore, the study demonstrates the model's robustness under different data distributions and varying numbers of nodes. This proves its effectiveness in real world applications, especially in multi-party collaboration contexts. Finally, the paper discusses challenges in optimizing blockchain performance and applying federated learning in privacy sensitive environments. It also outlines prospects for scalability and application in supply chain finance systems.