摘要
供应链金融是中小企业缓解融资困难的有效途径。为缓解供应链金融信用风险预测面临的信息不对称和样本选择偏差等问题,文中将供应链合作伙伴信息引入风险指标体系,基于2010-2021年A股上市企业披露数据,通过随机森林、XGBoost、逻辑回归和MLP四种机器学习模型进行比较分析。结果显示,合作伙伴信息的使用提高了供应链金融信用风险预测的准确性与稳定性。最后通过Lime可解释性分析,发现合作伙伴的速动比率、销售利润率、资产负债率等6个指标是供应链合作伙伴信息中影响信用风险预测的主要因素。
Supply chain finance(SCF)serves as an effective means for small and medium-sized enterprises(SMEs)to alleviate financing difficulties.To address challenges in predicting credit risks in SCF,such as information asymmetry and sample selection bias,this study incorporates supply chain partner information into the risk indicator system.Utilizing disclosed data from A-share listed companies between 2010 and 2021,a comparative analysis is conductedby by using four machine learning models:random forest,XGBoost,logistic regression,and MLP.The results demonstrate that the inclusion of partner information enhances the accuracy and stability of credit risk prediction in SCF.Finally,through Lime interpretability analysis,six indicators,including partner s quick ratio,sales profit margin,and asset-liability ratio,are identified as the primary influencing factors for credit risk prediction in supply chain partner information.
作者
张道海
杨晨
ZHANG Dao-hai;YANG Chen(School of Management,Jiangsu University,Zhenjiang 212013,China)
出处
《物流工程与管理》
2024年第4期56-60,共5页
Logistics Engineering and Management
基金
江苏省高校哲学社会科学研究重大项目(2020SJZDA063)
江苏省研究生科研与实践创新计划项目(SJCX22_1838)。
关键词
供应链金融
供应链合作伙伴
信用风险
机器学习
可解释性分析
supply chain finance
supply chain partners
credit risk
machine learning
interpretability analysis
