基于GMDH神经网络法的中小板上市公司ST分类预测研究  

Research on ST Classification Forecast of Small and Medium-sized Listed Companies Based on GMDH Neural Network Method

作　　者：石峰[1] SHI Feng(School of Management,Hunan Institute of Engineering,Xiangtan 411104,China)

机构地区：[1]湖南工程学院管理学院,湖南湘潭411104

出　　处：《湖南工程学院学报（社会科学版）》2021年第2期29-34,共6页Journal of Hunan Institute of Engineering(Social Science Edition)

基　　金：湖南省社科联项目“湖南小微企业债信评级模型构建及实证研究”(XSP17YBZZ023);湖南省情对策咨询项目“湖南中小企业债信评级体系及应用研究”(16JCC036);国家留学基金委地方合作项目“大学生返乡创业意愿及其提升路径研究”(留金项〔2018〕10006号)。

摘　　要：预测上市公司的上市状态对投资者与其他利益相关者至关重要。选取2019年1月1日至3月31日我国中小板116家上市公司的86个财务指标数据,运用GMDH神经网络法对ST公司和非ST公司进行分类预测,预测正确率达到78.26%。进一步采用1∶1配对原则(43家ST公司和43家非ST公司)对中小板上市公司进行ST分类预测,预测正确率大幅提高,达到88.24%。由此发现,基于GMDH神经网络法的中小板上市公司ST分类预测具有较好的预测表现。Predicting the listing status of listed companies is crucial for investors and other stakeholders.This paper selects 86 financial indicator data of 116 listed companies in China’s small and medium-sized board from January 1,2019 to March 31,2019,and uses GMDH neural network method to classify and predict ST companies and non-ST companies.The rate of prediction accuracy reaches 78.26%.Further adopting the 1∶1 matching principle(43 ST companies and 43 non-ST companies)to carry out ST classification prediction for small and medium-sized listed companies,the prediction accuracy rate has increased significantly,reaching 88.24%.The ST classification forecast of the listed company of GMDH neural network method far exceeds the conventional classification prediction method in the prediction accuracy.

关 键 词：GMDH神经网络法 中小板 ST公司 分类预测 

分 类 号：F275[经济管理—企业管理]