基于SMOTE-LR模型的上市公司失信风险评价研究

Research on the Evaluation of Listed Companies’Risk of Breach of Trust Based on SMOTE-LR Model

信用作为一切交易行为的根基,要求交易双方共同认定和遵守。只有交易双方践诺信用的基本准则,才能使双方实现无须付现的信用交易,从而促进经济高质量发展。近年来,在公司失信行为频发,并对社会造成恶劣经济影响的背景下,本文爬取了2014—2021年的上市公司失信数据作为研究对象,构建了上市公司失信风险评价模型,对公司失信风险进行研究。首先,本文获取了由财务指标和非财务指标构成的失信风险评价指标体系。其次,采用Lasso回归筛选出失信风险判别能力最优的指标体系。基于此,采用SMOTE算法处理不平衡样本,并结合逻辑回归算法构建SMOTE-LR模型对上市公司失信风险进行评估。输出的结果表明,除了传统财务指标之外,公司是否ST和审计意见等非财务指标也是影响公司失信风险的关键性因素。最后,采用更换机器学习的方式验证了本文SMOTE-LR模型的稳健性。由此,本文提出了三点政策建议。

Credit,as the foundation of all transaction behavior,requires both parties to the transaction to identify and comply with.Only when both parties to a transaction practice the basic guidelines of credit,it can enable both parties to realize credit transactions without paying cash,thus promoting high-quality economic development.In recent years,against the background of frequent corporate breach of trust and its bad economic impact on the society,this paper crawls the data of listed companies’breach of trust from 2014 to 2021 as the research object,constructs the evaluation model of listed companies’risk of breach of trust and conducts research on the risk of corporate breach of trust.Firstly,this paper obtains a credit default risk evaluation index system consisting of financial and non-financial indicators.Secondly,Lasso regression is used to screen out the index system with the best discriminatory ability of the default risk.Based on this,the SMOTE algorithm is used to process the unbalanced samples and the SMOTE-LR model is combined with the logistic regression algorithm to evaluate the default risk of listed companies.The output shows that in addition to traditional financial indicators,non-financial indicators such as whether the company is ST and audit opinion are also critical factors affecting the company’s risk of breach of trust.Finally,the robustness of the SMOTE-LR model in this paper is verified by using replacement machine learning.As a result,this paper puts forward three policy recommendations.