复杂金融担保网络的风险防控模型与算法研究

针对金融担保网络风险防控模型的研究，有助于预防大规模企业违约乃至系统性金融风险的发生，在保障国家金融安全、维护地方经济发展和稳定就业等方面有着重要作用。然而现有方法主要针对个体大中型企业、难以对大规模网络化的小微企业及其群体进行有效风险评估。针对上述问题，本申请拟探索复杂金融网络风险防控背后的核心数据科学理论问题，拟从三个方面开展研究工作：1）研究超约束异构网络表征空间中的个体风险评估；2）研究鲁棒高效的多源单向网络风险传播模型；3）以担保网络的整体和特定子图为目标，结合个体风险和传播风险，研究设计和开发系统性的担保网络风险防控模型。本项目拟通过个体、传播以及全局视角来综合评估网络中的信贷风险，研究用于智能监管的综合风险评估模型，探索数据科学驱动的金融网络风险防控中的核心科学问题。本项目立足于国家金融安全和智能监管的迫切需求，具有重要的理论意义和广阔的应用前景。

The risk evaluation research for networked-guarantee loans can effectively prevent the emergence and spread of the systemic credit crisis, which plays an important role in ensuring national financial security, maintaining economic development, and employment. However, the existing risk monitoring methods mainly target individual major enterprises. It is challenging to quantitatively analyze the risks of networked small and medium enterprises. To address the above challenges, this application explores the fundamental data science theory problem in risk management on complex networked-loans. It mainly includes 1) representative learning on constrained complex isomerism loan networks for individual risk assessment based on graph neural network; 2) research and design of robust and efficient network risk contagious model, which is multi-sources and directed; 3) combined with individual and contagion risk, aiming to explore the evaluation method of the systematic risk in loan network, design and develop a systematic early warning model of loan risk of guarantee network. This project intends to construct a systematic risk assessment model based on graph neural network, which carries out joint representations of individual, contagion, and systematic risk, so as to explore the fundamental theory innovation of data science for risk control on networked-loans. This project is based on the urgent needs of the nation's financial security and intelligent regulation, which will have a major impact on both the theoretical and application prospects.