基于联邦学习的风险管理研究进展  

Research Progress of Risk Management Based on Federated Learning

作　　者：戴雨璇 张永杰[1] 熊熊[1] 冯绪[1] 张维[1] Dai Yuxuan;Zhang Yongjie;Xiong Xiong;Feng Xu;Zhang Wei(College of Management and Economics,Tianjin University,Tianjin 300072,China)

机构地区：[1]天津大学管理与经济学部,天津300072

出　　处：《天津大学学报（社会科学版）》2022年第5期420-427,共8页Journal of Tianjin University：Social Sciences

基　　金：国家自然科学基金重点项目(92046024);国家自然科学基金重大项目(71790594).

摘　　要：近年来,随着数字经济的不断深入,在大数据的驱动下,人工智能、机器学习等取得了快速发展。数据的共享与流通成为了数字经济时代的必然要求,传统的数据服务模式也将发生改变。当前,数据科学的发展面临两大挑战,即数据的隐私保护与数据孤岛问题,联邦学习的出现恰好克服了以上两大挑战。联邦学习模型基于隐私保护实现数据共享的特性,引起了各个领域的关注并被认为具有重要的研究潜力。联邦学习算法和框架优化的相关研究在此后开展了最早的探索。文章通过文献综述的形式浅析联邦学习与区块链的结合、在医疗系统中的应用,梳理、总结了联邦学习的概念、分类、隐私保护技术、面临的挑战及其在小微企业中的应用情况。In recent years,with the continuous deepening of the digital economy,driven by big data,artificial intelligence and machine learning have achieved explosive developments.The sharing and circulation of data has become an inevitable requirement in the era of digital economy,and the traditional data service model will be changed.At present,the development of data science is faced with two major challenges,namely data privacy protection and data silos.The emerging federated learning overcomes the above two challenges.The federated learning model realizes data sharing based on privacy protection,which has attracted the attention of various fields and is considered to have important research potential.The related research on federated learning algorithms and framework optimization has been carried out since then.Through the literature review,the combination of federated learning and blockchain and its application in the medical system are analyzed.This paper also sorts out and summarizes the concept,classification,privacy protection technology,challenges faced and application in small and micro enterprises of federated learning.

关 键 词：联邦学习 数据孤岛 隐私保护 金融领域 风险评估 

分 类 号：F832.5[经济管理—金融学]