摘要
针对财务危机预警模型指标存在信息冗余及Logistic回归模型预测精度有待提高的不足,利用L1/2范数惩罚技术优化Logistic回归模型,构建基于L1/2正则化Logistic回归的上市公司财务危机预警新模型.通过以沪深股市制造业股票交易得到特别处理(Special Treatment, ST)公司和非ST公司为研究对象,对比研究传统Logistic回归和L1正则化Logistic回归模型的预测结果,实证研究表明:通过L1/2正则化的Logistic回归模型不仅可以实现参数估计和变量选择,而且具有更高的预测精度和泛化能力.研究体现了新模型对预警问题的合理性和优越性,为上市公司财务危机预警后续研究提供一定的借鉴.
There is information redundancy in the financial crisis early warning model and the prediction accuracy of Logistic regression model needs to be improved. In this paper,the L1/2 norm penalty technique is used to optimize the Logistic regression model, and a new financial crisis early warning model for listed companies based on L1/2 regularization Logistic regression is constructed. The special treatment(Special Treatment, ST) Company and non ST company as the research object was obtained by taking the stock trading of the Shanghai and Shenzhen stock market. The prediction result of the new Logistic regression modeling were compared with that of the traditional and L1 regularization Logistic regression models. The empirical study shows that the Logistic regression model of L1/2 regularization cannot only realize parameter estimation and variable selection but also has higher prediction accuracy and generalization ability. This study reflects the rationality and superiority of the new model to the early warning problem, and provides some reference for the following financial crisis early warning research of the listed companies.
作者
肖振红
杨华松
XIAO Zhen-hong;YANG Hua-song(Finance Department,Yunnan Agricultural University,Kunming 650201,China)
出处
《数学的实践与认识》
北大核心
2018年第21期80-89,共10页
Mathematics in Practice and Theory