摘要
针对企业财务风险研究指标数据中常常存在不平衡现象造成模型算法预测精度较差的问题,文章通过构造基于指导性正则化随机森林的SMOTEBoost算法分析企业财务风险因素,以此提升企业财务风险预测准确性,同时筛选出显著影响企业发生财务风险的特征变量。数值模拟结果发现SMOTEBoost-GRRF算法预测精度好于其他算法,且具有较优的特征变量筛选能力。实证研究结果发现每股资本公积金、营运资金、营业总成本/营业总收入、每股收益增长率四种指标是最能显著影响企业发生财务风险的因素。
Aiming at the problem that the prediction accuracy of the model algorithm is poor due to the unbalanced phenomenon in the index data of enterprise financial risk research,this paper analyzes the enterprise financial risk factors by constructing SMOTEBoost algorithm based on instructional regularized random forest,by which to improve the accuracy of enterprise financial risk forecasting.Meanwhile,the paper screens the characteristic variables that significantly affect the financial risk of enterprises.Numerical simulation results show that the SMOTEBoost-GRRF algorithm has better prediction accuracy than other algorithms,and it has better feature variable screening capabilities.The empirical results further confirms that capital accumulation fund per share,working capital,total operating cost/total operating revenue and growth rate of earnings per share are the four indexes that can most significantly affect the financial risk factors of enterprises.
作者
赵浩
鲁亚军
高洁
张汝飞
Zhao Hao;Lu Yajun;Gao Jie;Zhang Rufei(Postdoctoral Workstation of Credit Information Center of the People's Bank of China,Beijing 100031,China;Credit Information Center of the People's Bank of China,Beijing 100031,China;College of Economics and Trade,Hebei University of Geosciences,Shijiazhuang 050031,China)
出处
《统计与决策》
CSSCI
北大核心
2020年第4期9-14,共6页
Statistics & Decision
基金
河北省社会科学基金资助项目(HB17YJ024)。