主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2026, Vol. 34 ›› Issue (6): 13-21.doi: 10.16381/j.cnki.issn1003-207x.2024.2280cstr: 32146.14.j.cnki.issn1003-207x.2024.2280

• • 上一篇    下一篇

联邦学习与区块链赋能下的应收账款拍卖融资机制研究

刘琦铀, 陈嘉, 张成科, 朱怀念()   

  1. 广东工业大学经济学院,广东 广州 510520
  • 收稿日期:2024-12-16 修回日期:2025-04-18 出版日期:2026-06-25 发布日期:2026-05-22
  • 通讯作者: 朱怀念 E-mail:zhuhuainian@gdut.edu.cn
  • 基金资助:
    国家社会科学基金项目(22BGL067)

Research on Accounts Receivable Auction Financing under the Empowerment of Federated Learning and Blockchain

Qiyou Liu, Jia Chen, Chengke Zhang, Huainian Zhu()   

  1. School of Economics,Guangdong University of Technology,Guangzhou 510520,China
  • Received:2024-12-16 Revised:2025-04-18 Online:2026-06-25 Published:2026-05-22
  • Contact: Huainian Zhu E-mail:zhuhuainian@gdut.edu.cn

摘要:

供应链金融在农业领域中的应用面临诸多挑战,而数据共享与安全间的平衡难问题尤为突出。联邦学习与区块链赋能下的农业供应链融资虽融合了二者技术优势,但仍面临激励机制不足、计算效率较低等问题。有鉴于此,文章以应收账款融资业务为研究对象,首先构建了分层联邦学习与区块链赋能下的农业供应链融资系统。在此基础上,基于深度学习拍卖算法,将金融资源配置策略构建为以卖方收益最大化为目标的拍卖模型,以实现金融机构个体理性、激励相容性与合作社拍卖收入最大化。研究结果表明:通过拍卖策略优化,基于深度学习的拍卖机制能最大化卖方收益;合作社拍卖收入与其数据覆盖范围、金融机构数据覆盖范围要求成正比;神经网络近似优化率越高,系统优化难度越大,为此合作社需投入更多的计算资源和训练时间,进而影响其收益。

关键词: 应收账款拍卖融资, 区块链, 联邦学习, 深度学习

Abstract:

The insufficient penetration of full chain information and the difficulty in balancing data sharing and security make it difficult to effectively promote the application of supply chain financing in the agricultural field. The combination of blockchain and federated learning can fully leverage the technological advantages of both, effectively compensating for the shortcomings of blockchain in data privacy and security, and federated learning in data storage, synchronization, and tamper prevention. However, it still faces problems such as low computational efficiency and insufficient incentive mechanisms. Receivable financing business is taken accounts as the research object and an agricultural supply chain financing system empowered by hierarchical federated learning and blockchain is constructed; On this basis, based on deep learning auction algorithms, the financial resource allocation strategy is constructed as an auction model with the goal of maximizing seller returns, in order to achieve individual rationality, incentive compatibility, and maximization of cooperative auction revenue for financial institutions. Research has shown that: (1) through auction strategy optimization, auction mechanisms based on deep learning can maximize seller returns; (2) The auction revenue of cooperatives is directly proportional to their data coverage and the data coverage requirements of financial institutions; (3) The higher the approximate optimization rate of neural networks, the greater the difficulty of system optimization. Therefore, cooperatives need to invest more computing resources and training time, which in turn affects their profits. New ideas for promoting the application of digital intelligence technology in the field of agricultural supply chain financing are provided, and certain practical significance is given in promoting data information security and sharing, fund integration and optimization, and assisting rural industrial revitalization.

Key words: accounts receivable auction financing, blockchain, federated learning, deep learning

中图分类号: