中国企业“僵尸化”的SMOTE-SVM智能预警研究

Research on Intelligent Early Warning for Chinese Enterprises Zombified Based on SMOTE-SVM

以中国上市公司为研究对象,从流动性状况、偿债能力状况、盈利能力状况和市场价值分析四方面财务指标提取20项预警指标变量;依据僵尸企业衍变过程对"财务危机"进行界定,进而获得财务危机预警建模的状态指标变量,并构建SMOTE-SVM模型。实证结果表明:加入SMOTE的预警模型较未加入SMOTE的预警模型具有更高的预测精度;同时,在基于各类样本的情况下,加入SMOTE的SVM预警模型均优于加入了SMOTE的其余预警模型。此外,基于不同区域企业样本所构建的SMOTE-SVM较基于整体企业样本以及基于不同所有制企业样本构建的SMOTE-SVM均具有更为优越的预测性能。

Taking Chinese listed companies as the research object,four financial indicators from the liquidity situation,the solvency status,the profitability and the market value analysis are selected and extracted as 20 early warning index variables in this paper.Meanwhile,the'financial crisis'is defined based on the evolution of the zombie enterprises and the state index variable of the financial crisis early warning model is obtained,and built SMOTE-SVM model.The empirical results show that the early warning model with SMOTE has a higher prediction accuracy than that of the early warning model without SMOTE.At the same time,the early warning model based on SVM with SMOTE is better than the other early warning models with SMOTE.In addition,SMOTE-SVM based on enterprise samples from different regions has better prediction performance than that based on whole enterprise samples and different ownership enterprise samples.