基于集成算法的上市公司财务困境预警模型  

The Study on Financial Distress Prediction of Listed Companies Based on Ensemble Model

作　　者：陈辉远 姜慜喆 冯家兴 CHEN Huiyuan;JIANG Minzhe;FENG Jiaxing(School of Safety Science and Emergency Management,Wuhan University of Technology,Wuhan 430070,China;不详)

机构地区：[1]武汉理工大学安全科学与应急管理学院,湖北武汉430070 [2]华中科技大学管理学院,湖北武汉430070

出　　处：《武汉理工大学学报（信息与管理工程版）》2022年第3期468-477,共10页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：国家重点研发计划课题(2018YFC080700)。

摘　　要：为保证上市公司财务困境预警的准确性,结合财务指标和管理指标构建上市公司财务困境预警指标体系,提出基于集成算法的财务困境预警模型,利用SMOTE过采样技术处理非平衡数据,并采取遗传算法进行特征选择,通过对比分析多种单分类器模型与集成分类器模型在平衡数据集与非平衡数据集下的表现,选出更适用于财务困境预警指标体系的模型构建方式。结果表明:对非平衡数据集进行处理可以大幅度提高模型性能,增强上市公司对财务困境的预警能力;相比单分类器,集成分类器在上市公司财务困境预警问题上具有更好的表现。In order to ensure the accuracy of financial distress prediction of listed companies, this paper combined financial indicators and management indicators to build the financial distress early warning index system of listed companies, to develop the financial distress early warning model based on ensemble model, use SMOTE oversampling technology to deal with unbalanced data, and adopt genetic algorithm for feature selection. By comparing and analyzing the performance of multiple single classifier models and ensemble classifier models under balanced data sets and unbalanced data sets, we selected the model construction method which is more suitable for financial distress prediction. The experimental results show that:(1) Processing unbalanced data sets can greatly improve the performance of the model, and improve the ability of listed companies to predict financial distress;(2) Compared with single classifier, integrated classifier has better performance on financial distress warning of listed companies.

关 键 词：非平衡数据集 财务困境预警 管理指标 集成算法 遗传算法 

分 类 号：G356[文化科学—情报学]