现代电子技术2024，Vol.47Issue(8)：43-48,6.DOI:10.16652/j.issn.1004-373x.2024.08.007
基于多目标优化加权软投票集成算法的信用债违约预警研究
Research on credit bond default warning based on multi objective optimization weighted soft voting integration algorithm
摘要
Abstract
In order to enhance the accuracy and stability of credit bond default prediction for the purpose of financial risk management,a multi-objective optimized weighted soft voting ensemble algorithm is proposed by taking the credit bonds(January 1,2014,to December 31,2021)as the object of study.In this algorithm,the recognition capability of each base classifier is quantified by calculating their fuzzy densities,and the multi-objective particle swarm optimization algorithm is used to slove the weights of the base classifiers.In comparison with other individual classifiers such as support vector machine,logistic regression,Gaussian naive bayes,multi-layer perceptron(MLP),as well as other ensemble algorithms like voting and stacking,feature importance analysis is conducted by means of the expected permutation feature importance(PFI)algorithm.The results indicate that the weighted soft voting ensemble algorithm exhibits outstanding performance in credit bond default prediction.It not only enhances the performance of individual algorithms but also demonstrates higher accuracy,precision,and AUC values compared to other ensemble algorithms.Features with higher importance in credit bond default prediction include the issuer's credit rating,exchange,bond rating prior to default,total asset turnover ratio,monetary funds,net asset growth rate,operating cash flow as a percentage of revenue,GDP,PPI,registered location,short-term government bond interest rates,leading economic indicators,bond type,and industry sector.This research can provide an effective approach for financial risk prediction,offering valuable insights for investors and financial institutions in the risk warning.关键词
金融风险管理/信用债违约预警/加权软投票集成算法/多目标优化/模糊密度/期望PFI算法Key words
financial risk management/credit bond default warning/weighted soft voting ensemble algorithm/multi objective optimization/fuzzy density/expected PFI algorithm分类
电子信息工程引用本文复制引用
郑怡昕,王重仁..基于多目标优化加权软投票集成算法的信用债违约预警研究[J].现代电子技术,2024,47(8):43-48,6.基金项目
山东省科技型中小企业创新能力提升工程(2023TSGC0208) （2023TSGC0208）
