摘要
采用Boruta算法选择中小企业供应链金融信用风险指标原始的特征变,挑选出相对重要的特征变量.运用Stacking集成算法,融合多种单一预测模型,构建Boruta-Stacking集成信用风险评估模型.并将该模型运用于计算机、通信和其他电子设备制造业中小企业的供应链融资和信用评级.实验结果表明,所建立模型预测的准确性达到97.14%,高于单一模型的预测准确性,并使用部份依赖图(PDP)揭示重要特征变量与中小企业信用风险之间的关系.
The Boruta algorithm is used to select the original characteristic variables of the credit risk index of the supply chain finance of SMEs,and the relatively important characteristic variables are selected.On this basis,the Stacking integrated algorithm is used to integrate multiple single prediction models to construct a Boruta-Stacking integrated credit risk assessment model.The model is also applied to supply chain financing and credit ratings for SMEs in the computer,communications and other electronics manufacturing industries.Experimental results show that the prediction accuracy of the established model reaches 97.14%,which is higher than that of a single model,and partial dependency graphs(PDPs)is used to reveal the relationship between important feature and credit risk of SMEs.
作者
徐超强
李碧珍
XU Chaoqiang;LI Bizhen(School of Economics,Fujian Normal University,Fuzhou Fujian 350001 China;School of Economics and Law,Concord University College Fujian Normal University,Fuzhou Fujian 350001 China)
出处
《泉州师范学院学报》
2023年第2期63-70,共8页
Journal of Quanzhou Normal University
基金
福建省科技厅创新战略研究项目(2020R0045)
福建省高校哲学社会科学研究项目(统一战线专项)(JAT21054)。
作者简介
通信作者:李碧珍(1966-),女,福建泉州人,教授,博士,从事企业经济研究,E-mail:451299537@qq.com.
