基于Twin SVR的公司违约风险预测研究

›› 2019, Vol. 31 ›› Issue (11): 33-43.

基于Twin SVR的公司违约风险预测研究

林宇¹, 吴庆贺¹, 李昊^1,2, 唐晓华³

1 成都理工大学商学院, 成都 610059;
2 中国农业银行合肥新站高新区支行, 合肥 230011;
3 四川工商职业技术学院, 成都 611830

收稿日期:2019-03-21 出版日期:2019-11-28 发布日期:2019-11-30
通讯作者: 林宇,成都理工大学商学院教授,博士生导师,博士
作者简介:吴庆贺,成都理工大学商学院硕士研究生;李昊,中国农业银行合肥新站高新区支行,硕士;唐晓华,四川工商职业技术学院讲师,硕士。
基金资助:
国家自然科学基金面上项目（71771032）；四川省科技计划项目（2017JY0158）；四川矿产资源研究中心资助项目（SCKCZY2016-ZC05）。

Research on Corporate Default Risk Prediction Based on Twin SVR

Lin Yu¹, Wu Qinghe¹, Li Hao^1,2, Tang Xiaohua³

1 School of Business, Chengdu University of Technology, Chengdu 610059;
2 Gaoxin Sub-Branch, Hefei Branch, Agriculture Bank of China, Hefei 230011;
3 Sichuan Technology Business College, Chengdu 611830

Received:2019-03-21 Online:2019-11-28 Published:2019-11-30

摘要/Abstract

摘要：

本文以发行短期融资券的公司为研究对象，首先基于信用利差构建出公司违约风险变量，然后运用Twin SVR模型对公司违约风险展开预测，最后对影响公司违约风险Twin SVR模型预测性能的特征指标解释能力进行了探讨。实证结果表明：在Twin SVR模型的构建过程中，发现RBF核函数展示出了比Polynomial、Linear和Sigmoid核函数更加卓越的预测性能；与传统的SVR、BPNN以及Logistic公司违约风险预测模型相比，Twin SVR模型不仅在整体行业中具有最优的预测性能，而且在分行业中具有优异的泛化性能；对于影响Twin SVR模型预测性能的前两位特征指标解释能力来说，在整体行业和有色金属行业中前两位最具解释能力的指标为信用评级和是否为国有企业，在煤炭和钢铁行业中为信用评级和净资产收益率。

关键词: Twin SVR, 公司违约风险, 短期融资券, 预测

Abstract:

This paper, taking companies that issue short-term financing bond as the objects of research, establishes a suitable method based on the credit spread to analyze the corporate default risk, then applies Twin-SVR (Twin Support Vector Regression) to predict the corporate default risk and finally explores the ability to interpret the characteristic indicators that affect the predictive performance of Twin-SVR default risk model. The empirical results are as follows:During the construction of the Twin-SVR model, we find that the RBF kernel function exhibits better prediction performance than the Polynomial, Linear, and Sigmoid kernel functions; compared with the traditional SVR, BPNN and Logistic corporate default risk prediction models, the Twin-SVR model not only has the best prediction performance and stability performance in the overall industry, but also has excellent generalization performance in the sub-industry; for the interpretation ability of the top two characteristic indicators that affect the prediction performance of Twin-SVR model, we find that the most explanatory two indicators in the overall industry and non-ferrous metal industry are credit rating and SOE/non-SOE background, and in the coal and steel industry, the top two indicators are credit rating and ROE.

Key words: Twin SVR, corporate default risk, short-term financing bond, prediction

林宇, 吴庆贺, 李昊, 唐晓华. 基于Twin SVR的公司违约风险预测研究[J]. , 2019, 31(11): 33-43.

Lin Yu, Wu Qinghe, Li Hao, Tang Xiaohua. Research on Corporate Default Risk Prediction Based on Twin SVR[J]. , 2019, 31(11): 33-43.

参考文献

[1] 陈艺云.中国公司债违约风险度量的理论与实证研究[J].系统工程, 2016,34(1):26-33
[2] Huang X. B., Liu X. L., Ren Y. Q. Enterprise Credit Risk Evaluation Based on Neural Network Algorithm[J]. Cognitive Systems Research, 2018,52:317-324
[3] Sohn S. Y., Kim D. H., Yoon J. H. Technology Credit Scoring Model with Fuzzy Logistic Regression[J]. Applied Soft Computing, 2016,43:150-158
[4] Zhang Y., Shi B. Non-tradable Shares Pricing and Optimal Default Point Based on Hybrid KMV Models:Evidence from China[J]. Knowledge-Based Systems, 2016,110:202-209
[5] Yang Y., Gu J., Zhou Z. F. Credit Risk Evaluation Based on Social Media[J]. Environmental Research, 2016,148:582-585
[6] 曹麟.监管资本与经济资本存在本质区别吗?[J].上海金融, 2014,(2):63-68
[7] Tam K. Y., Kiang M. Y. Predicting Bank Failures:A Neural Network Approach[J]. Applied Artificial Intelligence, 1990,4(4):265-282
[8] 林宇,黄迅,淳伟德,等.基于ODR-ADASYN-SVM的极端金融风险预警研究[J].管理科学学报, 2016,19(5):87-101
[9] Vapnik V. N. The Nature of Statistical Learning Theory[M]. New York:Springer-Verlag, 1995
[10] 陈垚,毛保华,柏赟,等.基于支持向量回归的地铁牵引能耗预测[J].系统工程理论与实践, 2016,36(8):2101-2107
[11] 王鹏,黄迅.基于Twin-SVM的多分形金融市场风险的智能预警研究[J].统计研究, 2018,35(2):3-13
[12] Peng X. J. TSVR:An Efficient Twin Support Vector Machine for Regression[J]. Neural Networks, 2010,23(3):365-372
[13] Wang H. M., Xu Y. T. Scaling Up Twin Support Vector Regression with Safe Screening Rule[J]. Information Sciences, 2018,465:174-190
[14] 曹勇,李孟刚,李刚,等.基于信用利差与Logistic回归的公司违约概率测算模型与实证研究[J].运筹与管理, 2016,25(6):209-223
[15] 钟灿辉,陈武.银行信贷实务与管理[M].成都:西南财经大学出版社, 2005
[16] 彭爽,李永清.我国商业银行发展非金融企业短期融资券业务研究[J].经济问题, 2017,(7):64-68
[17] 宫晓莉,庄新田.双指数跳跃扩散条件下上市公司违约风险分析[J].系统工程学报, 2018,33(1):44-54
[18] 淳伟德,肖杨.供给侧结构性改革期间系统性金融风险的SVM预警研究[J].预测, 2018,37(5):36-42
[19] 吴建华,张颖,王新军.信息披露扭曲下企业债券违约风险量化研究[J].数理统计与管理, 2017,36(1):175-190
[20] 陈诗一.德国公司违约概率预测及其对我国信用风险管理的启示[J].金融研究, 2008,(8):53-71
[21] 段文奇,宣晓.基于价值创造视角的互联网企业价值评估体系研究[J].财贸研究, 2018,29(9):85-97
[22] 迟国泰,潘明道,程砚秋.基于综合判别能力的农户小额贷款信用评价模型[J].管理评论, 2015,27(6):42-57
[23] 申敏,吴和成.国民经济行业信用风险相依结构分析[J].软科学, 2016,30(3):126-129
[24] 沈中华,管超,毕胜,等.信用评级有效性研究——发行利差的解构视角[J].吉林大学社会科学学学报, 2018,58(6):94-105
[25] 顾晓安,彩玲.产能过剩、企业经营风险与地区金融安全——来自山西省煤炭上市公司的经验证据[J].当代经济管理, 2018,40(8):70-75