基于COX模型的财务危机动态预测方法

企业爆发财务危机是一个动态的演化过程,与财务危机静态预测研究不同的是,文章采用多元时间序列数据表述企业特征,并结合COX模型进行动态预测研究。从多角度采集企业爆发财务危机的影响变量,完成必要的数据预处理后采用Mann-Whitney U检验和多重共线性检验进行变量筛选,并提取财务危机预测的主成分变量以降低变量维度,最后对比分析静态COX模型、动态COX模型和其它模型的预测效果。沪深两市A股上市公司的全样本实证结果表明:动态COX预测模型的综合识别准确率比静态COX预测模型高,且分类识别的第I类准确率和第II类准确率均有所提高,与其它模型相比具有较好的预测效力。动态COX预测模型具有较好的财务危机预判能力。

Dynamic Forecasting Method of Financial Crisis Based on COX Model

The outbreak of financial crisis is a dynamic evolution process. Different from the static prediction research of financial crisis, this paper uses multivariate time series data to express the characteristics of enterprises, and carries out dynamic prediction research combined with Cox model. It collects the influencing variables of financial crisis from multiple perspectives, complete the necessary data preprocessing, use Mann Whitney U test and multicollinearity test to screen variables, and extract the principal component variables of financial crisis prediction to reduce the variable dimension. Finally, it analyzes the prediction effect of static Cox model, dynamic Cox model and other models. The full sample empirical results of A-share listed companies in Shanghai and Shenzhen stock markets show that the comprehensive recognition accuracy of dynamic Cox prediction model is higher than that of static Cox prediction model, and the class I and class II accuracy of classification recognition are improved, which has better prediction effect compared with other models. Dynamic Cox prediction model has a good ability to predict financial crisis.