基于主成分分析和Logistic回归的财务风险预警研究

Research on Early Warning of Financial Risk Based on Principal Component Analysis and Logistic Regression

摘要: 本文选取了沪深A股市场中的100家上市公司为样本，选取了包括20个财务预警指标和2个非财务预警指标。通过主成分分析将20个财务预警指标提取7个主成分因子，并将7个主成分因子和2个非财务预警指标作为自变量，企业类型为因变量利用python进行二元logistic回归建模和实证性检验，取得了比较理想的预测结果，且模型优劣评价的AUC值超过0.85，KS值达到0.67，证实了模型的预测结果具有较高的可靠性，表明该模型能够为企业经营者及时发现财务风险隐患，调整经营策略提供一定的实用价值。

Abstract: In this paper, 100 listed companies in Shanghai and Shenzhen A-share market are selected as samples, including 20 financial early warning indicators and 2 non-financial early warning indicators. The 20 financial warning indicators were extracted from 7 principal component factors through principal component analysis, and the 7 principal component factors and 2 non-financial warning indicators were used as independent variables, and the type of enterprise was used as the dependent variable to carry out binary logistic regression modeling and empirical testing using python, and the prediction results were satisfactory, and the AUC value of the model’s merit evaluation exceeded 0.85, and the KS value reached 0.67, which confirms the high reliability of the model’s prediction results, indicating that the model can provide certain practical value for business operators to discover financial risks in time and adjust business strategies.