基于贝叶斯网络的上市证券公司风险预警模型研究
    点此下载全文
引用本文:聂瑞华1,石洪波2.基于贝叶斯网络的上市证券公司风险预警模型研究[J].财经理论与实践,2018,(6):51-57
摘要点击次数: 832
全文下载次数: 0
作者单位
聂瑞华1,石洪波2 (1.山西财经大学 统计学院山西 太原 0300062.山西财经大学 信息与管理学院山西 太原 030006) 
中文摘要:上市证券公司的风险预警模型能够为政府监管、证券公司稳健发展以及投资者研判提供依据。以上市证券公司风险管理指标体系为基础,利用贝叶斯网络方法以及支持向量机、随机森林和多项Logit模型分别建立风险预警模型进行比较,并在实证中针对上市证券公司的不平衡数据特征,用 SMOTE抽样对数据进行预处理。最终实证表明:从平均准确率和标准差两个角度比较,SOMTE抽样增加了贝叶斯网络的预测效果,机器学习方法要优于多项Logit模型,贝叶斯网络方法效果最佳。
中文关键词:风险预警模型  贝叶斯网络  SMOTE抽样
 
A Study on Risk Warning Model of Listed Securities Companies Based on the Bayesian Network
Abstract:Risk early-warning models of listed securities companies provide the basis for government regulation, stable development of listed securities companies and conscious judgments of investors. Based on established risk management index system of listed securities companies, this paper uses Bayesian Network, support vector machines, random forest and multinomial logit regression to set up risk early-warning models separately. And in the empirical analysis, the data are pre-treated by SMOTE sampling in view of the unbalanced data characteristics of listed securities companies. The empirical results show that the SOMTE sampling increases the prediction effect of the algorithm from the 2 angles of the average accuracy rate and the standard deviation. The prediction model established by the machine learning algorithms are obviously better than the multinomial logit model. And the Bayesian network is the best.
keywords:risk warning model  Bayesian network  SMOTE sample
查看全文   查看/发表评论   下载pdf阅读器