基于GMDH与dce-GMDH算法的上市公司财务危机预测研究

Research on Financial Crisis Prediction of Listed Companies Based on GMDH and dce-GMDH algorithm

预测上市公司的上市状态有助于了解企业的经营状态、稳定市场预期。选取2019年第4季度沪深A股上市公司的138家ST公司和138家非ST公司样本组,并根据盈利能力、经营增长、资产质量和债务风险等4个维度的17个财务指标,运用GMDH算法和基于多种分类器集合的GMDH(dce-GMDH)算法对ST公司和非ST公司进行财务危机分类预测,结果表明:两种算法在ST公司和非ST公司两类公司中取得较高的分类预测效果。GMDH算法的预测准确率为83.64%,dce-GMDH算法的预测准确率为85.45%。与GMDH算法相比,使用dce-GMDH算法可将分类预测准确率提高约2%。

Predicting the listing status of listed companies helps to understand the business status of the company and stabilize market expectations.This paper selects 138 ST companies and 138 non-ST companies as sample groups of China′s Shanghai and Shenzhen A-share listed companies in the first quarter of 2019,and 17 financial indicators based on four dimensions:profitability,business growth,asset quality and debt risk.The GMDH algorithm and the dce-GMDH are used to predict the financial crisis classification of ST companies and non-ST companies.Both algorithms achieve higher prediction accuracy.The prediction accuracy of GMDH algorithm is 83.64%,and the prediction accuracy of dce-GMDH algorithm is 85.45%.Compared with the GMDH algorithm,the dce-GMDH algorithm can improve the classification prediction accuracy by about 2%.