| 摘要: |
| 针对传统预测模型对于企业信用债券违约预测准确率低、拟合效果差的问题,提出了基于Kaufman-Merton-Voss （KMV）-Categorical Boosting （CatBoost）的企业债券违约预测模型. 首先对原始样本数据进行预处理,降低噪声数据对预测模型的影响；然后,利用KMV模型评估借款公司信用违约概率,计算公司资产市场价值与公司资产市场价值的波动率,获得企业资产价值与违约点之间的差额Distance-to-Default（DD）；最后,利用债务偿还期限、短期无风险收益率、公司股权市场价值、公司债务面值计算出的违约距离,将其加入指标中,利用CatBoost算法预测企业信用债券违约风险,通过基于Ordered Boosting方式的CatBoost算法训练模型,得到无偏梯度估计,以减缓预测偏移,从而增强模型的泛化能力. 实验结果表明：基于KMV-CatBoost增强的模型能够提高企业信用债券违约风险识别的准确率,识别正确率约为95.5%. |
| 关键词: 债券违约 预测模型 CatBoost Kaufman-Merton-Voss(KMV) |
| DOI：10.3969/J.ISSN.1000-5137.2024.02.016 |
| 分类号:TP183 |
| 基金项目:上海市科学技术委员会项目（22142201900） |
|
| An identification of default risk in corporate credit bonds based on KMV-CatBoost enhanced model |
|
WANG Peipei, ZHOU Xiaoping, CHEN Jiajia, WANG Hanqi
|
|
College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 201418, China
|
| Abstract: |
| Aiming at the problems of low accuracy and poor fitting effect of traditional prediction models for corporate credit bond default prediction, a corporate bond default prediction model based on Kaufman-Morton-Voss （KMV）-Categorical Boosting （CatBoost） was proposed. Firstly, the original sample data was preprocessed to reduce the impact of noisy data on the prediction model. Secondly, the credit default probability of the borrowing company was evaluated by using the KMV model. The market value of the company’s assets and the volatility of it were calculated, in order to obtain the Distance-to-Default （DD） difference between the company’s asset value and the default point. Finally, the default distance was calculated by adding debt repayment period, short-term risk-free return rate, company equity market value, and company debt face value to the indicators. The CatBoost algorithm was used to predict the default risk of corporate credit bonds. The model was trained using the Ordered Boosting based CatBoost algorithm to obtain unbiased gradient estimation, which slowed down prediction bias and enhanced the model’s generalization ability. The experimental results showed that the KMV CatBoost enhanced model could improve the accuracy of identifying default risk in corporate credit bonds, with a recognition accuracy of approximately 95.5%. |
| Key words: bond default prediction model CatBoost Kaufman-Merton-Voss (KMV) |
