检索规则说明:AND代表“并且”;OR代表“或者”;NOT代表“不包含”;(注意必须大写,运算符两边需空一格)
检 索 范 例 :范例一: (K=图书馆学 OR K=情报学) AND A=范并思 范例二:J=计算机应用与软件 AND (U=C++ OR U=Basic) NOT M=Visual
作 者:迟国泰[1] 王珊珊[1] 王逸然 CHI Guotai;WANG Shanshan;WANG Yiran(School of Economics and Management,Dalian University of Technology,Dalian 116024,China)
出 处:《系统工程理论与实践》2025年第2期481-502,共22页Systems Engineering-Theory & Practice
基 金:辽宁省社会科学规划基金(L21BGL011)。
摘 要:违约预测作为金融机构区分潜在违约借款人的有效工具,一直被用于信用风险评估领域.针对传统Stacking方法中元分类器较弱、预测效果不理想的现象,构建基于Stacking集成学习的违约风险预警模型.根据多种基准模型对比思路,在准确率等六个模型评价指标上,验证Stacking模型对中国A股3425家上市公司违约风险预警的有效性;基于德国等五个公开信用数据集,通过Friedman检验与Bonferroni-Dunn检验,验证本文模型的稳健性.创新与特色:一是通过Lasso-logistic模型,在众多指标组合中,遴选违约鉴别能力最大的指标组合.二是基于不同的基分类模型组合确定最优元分类器,构建Stacking集成学习模型预测上市公司违约风险,证明了不同方法的模型组合可以提高违约预警模型的分类性能,为公司信用风险评价研究提供新思路.研究表明:采用最优元分类器构建Stacking模型的违约预测精度F-measure有所提高;在多个评价指标上,Stacking模型的预测性能普遍优于逻辑回归、决策树等多种基准模型;“带息债务/全部投入资本”“货币资金比例”及“审计意见类型”等指标对预测公司未来1~5年的违约风险具有重要作用.Default prediction has become an efficient tool that allows financial institutions to differentiate their potential default borrowers,which has been applied in credit risk assessment.Due to the drawbacks of weak meta-classifier and poor predictive ability in traditional Stacking method,this study constructs a default risk warning model based on Stacking approach.Based on the motivation of multiple benchmark model comparisons,the proposed model’s efficiency is confirmed from the perspective of six different performance measures including accuracy with respect to forecasting the default risk of 3425 Chinese A-share listed companies.Moreover,we use Friedman test and Bonferroni-Dunn test to verify the robustness of the proposed model based on five open credit datasets including German.There are two innovations and features in this study.First,the optimal feature set is obtained among many feature sets using Lasso-logistic model.Secondly,this study establishes a Stacking ensemble learning model that determines the optimal meta-classifier based on different base classification model combinations for warning the default risk of listed companies,which contributes to the field of credit scoring research by demonstrating that model combinations of different methods are worth considering to improve the classification performance of default prediction models.Our experimental results demonstrate that F-measure of the proposed model constructed based on the optimal meta-classifier has improved.In terms of multiple performance measures,the proposed model’s predictive performance outperforms several other benchmark models including logistic regression and decision tree.These features,including interest-bearing debt/total invested capital,monetary fund ratio,and type of audit opinion,play an important role in forecasting the default risk of a company in the next 1∼5 years.
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在链接到云南高校图书馆文献保障联盟下载...
云南高校图书馆联盟文献共享服务平台 版权所有©
您的IP:216.73.216.38