摘要
为评估各因素对净资产收益的影响程度,以2002—2022年我国上市煤炭企业的相关数据为研究样本,基于机器学习中的Bayesian-Ridge模型进行训练,分析各自变量对因变量的影响程度,并与传统OLS模型进行对比。实验结果表明:在影响因素权重大小方面,Bayesian-Ridge模型与OLS模型均显示经营负债比率对净资产收益率影响最大,其次是国有持股比例和营业利润率;在方法层面,随样本点数据增加,Bayesian-Ridge模型可降低先验分布的影响;Bayesian-Ridge模型决定系数R~2为0.002 3,高于OLS模型的0.001 8,在该问题研究中Bayesian-Ridge模型优于OLS模型。
This paper is aimed at an evaluation of the impact of various factors on net asset returns.Taking the relevant data of listed coal enterprises in China from 2002 to 2022 as the research sample,the study involves training the Bayesian-Ridge model in machine learning,analyzing the impact of each independent variable on dependent variable,and comparing it with traditional OLS model.The experimental results show that in terms of the weight of influencing factors,both Bayesian-Ridge model and OLS model show that the operating debt ratio has the greatest impact on the return on equity,followed by the state-owned shareholding ratio and operating profit margin;and at the method level,as the sample point data increases,Bayesian-Ridge model can reduce the influence of prior distribution;the R~2 value of the Bayesian-Ridge model is 0.002 3,which is higher than R~2 value of the OLS model by 0.001 8.Thus,Bayesian-Ridge model is superior to OLS model.
作者
谭旭红
王朕卿
Tan Xuhong;Wang Zhenqing(Heilongjiang University of Science&Technology,Harbin 150022,China;School of Management,Heilongjiang University of Science&Technology,Harbin 150022,China)
出处
《黑龙江科技大学学报》
CAS
2023年第4期622-628,共7页
Journal of Heilongjiang University of Science And Technology
作者简介
第一作者:谭旭红(1968-),女,黑龙江省勃利人,教授,博士后,研究方向:企业管理,E-mail:tanxuhong@126.com。
