供应链金融风险评估研究综述——基于知识图谱技术

A review of supply chain finance risk assessment research:Based on knowledge graph technology

供应链金融作为一种面向供应链上所有成员企业的系统性金融服务,重在解决中小微企业融资难、融资贵问题.但长期以来相关参与主体信息不对称问题导致其风险频发,特别是风险评估技术相对滞后,使得供应链金融难以得到有效推广.因此,供应链金融风险评估成为业界亟需解决和学界广泛研究的热点问题.本文主要从供应链金融风险评估相关文献着手,通过知识图谱技术从152篇中文文献和61篇英文文献中提取知识,并进行可视化分析.一方面探究了供应链金融风险评估的研究现状,发现现有研究存在研究视角片面、研究对象单一、研究数据样本小、研究模型性能和可解释性不足等问题;另一方面展望供应链金融风险评估的研究趋势,主要包括以下问题有待深入探究:如何多渠道采集结构化、半结构化和非结构化数据,并对其进行有效整合,以及实现隐私数据的信息安全共享.如何从互联网经济活动所产生的海量数据中挖掘供应链金融风险关联知识,并从中提取风险特征,以此建立高维度、细粒度的供应链金融风险评估指标体系.如何运用深度学习方法来提升供应链金融风险评估模型的整体性能,并通过解析风险特征的重要性和边际效应来确保模型的可解释性.

Supply chain finance(SCF),as a systematic financial service for all member enterprises in the supply chain,focuses on solving the financing problems of small and medium-sized enterprises(SMEs).However,for a long time,the asymmetric information of relevant participants leads to frequent risks,and the risk assessment technology is relatively backward,which makes it difficult to effectively promote the service model.Therefore,the assessment of SCF risk has become a hot issue in the industry and academic circles.In this paper,the related documents of SCF risk assessment are taken as the research object,and knowledge is extracted from 152 Chinese documents and 61 English documents by using knowledge graph technology,and then visual analysis is carried out.On the one hand,the current research status of SCF risk assessment is explored,and it is found that the existing research has one-sided research perspective,single research object,small research data sample,and insufficient research model performance and interpretability.On the other hand,looking ahead to the research trends in SCF risk assessment,the main issues include the following to be explored in depth:1)How to collect structured,semi-structured and unstructured data from multiple sources,integrate them effectively,and share private data securely?2)How to mine the SCF risk association knowledge from the massive data generated by Internet economic activities and extract risk features from it,to establish a high-dimensional and fine-grained SCF risk assessment index system?3)How to use deep learning methods to improve the overall performance of SCF risk assessment models and ensure the interpretability of the models by resolving the importance and marginal effects of risk features?