基于平均准确度和贝叶斯准则指标遴选的企业违约预测研究——以中国上市小企业为例

摘要/Abstract

摘要： 研究上市小企业违约风险对缓解小企业融资难、融资贵难题以及促进经济平稳发展具有重要意义。本研究基于大数据视角从上市小企业财务指标、非财务指标和宏观经济指标等300多个指标中遴选出一组最优指标组合，并用这个最优指标组合建立中国上市小企业违约预测模型。本文的创新和特色:以指标X_i的基尼系数最小作为划分违约上市小企业和非违约上市小企业的节点。以一个指标组合构造一棵决策树，从m个指标中采用随机抽样的方法得到多个指标组合，并用这多个指标组合构建多棵决策树。计算任意打乱指标X_i的取值前后多棵决策树的违约预测误差值的差值均值，作为指标的重要性值VI(X_i)，并保留重要性较大的指标。以保留的指标和上市小企业违约状态构造贝叶斯准则值，以每次增加一个指标和减少一个指标的方法，遴选贝叶斯准则值最小时的指标组合作为最优指标组合。研究表明:(1)本研究遴选的指标能预测上市小企业违约状态。(2)多数据集对比分析、多指标遴选对比分析和多模型对比分析结果表明，本文指标遴选方法优于蓝本文献和其他3种典型的指标遴选方法。(3)不同地理区域的上市小企业信用资质最高的是东部地区。不同上市途径的上市小企业信用资质最高的是非借壳上市小企业。

关键词: 平均准确度, 企业违约, 违约预测, 指标遴选

Abstract: Predicting potential default of small listed companies (small listcos for short) is of great significance for banks to issue loans, for investors to make an informed decision and for regulatory authorities to supervise the market. Based on the big data perspective, this paper selects a set of indicators from more than 300 indicators including small listcos' financial and non-financial indicators and macro-economic indicators, and uses the set of indicators to establish a default prediction model for China's small listcos. The contributions of this paper are as follows:in the first stage, the minimum value of Gini of the indicator Xi is used as the node to divide the small listcos into defaulting ones and non-defaulting ones; a set of features is used to construct a decision tree and then multiple sets of indicators are used to construct multiple decision trees, which are used to calculate the importance value VI (Xi), with indicators of high importance value retained; a set of indicators with the minimum Bayesian criterion value are selected from the retained indicators. The research shows that:(1) The indicators selected in this study can predict the default status of small listcos and conform to the principle of credit 5C. The indicator that has a key impact on the default status of small listcos in the next 0-5 years is “retained earnings per share”; the indicator that has a key impact on the short-term default forecast of small listcos is “other receivables”; an indicator that has a key im-pact on default forecasts is the “Industry Sentiment Index”; (2) the comparative analysis of multiple data sets, the comparative analysis of multi-index selection and the comparative analysis of multiple models show that the index selection method in this paper is significantly better than other three typical methods, and the results are robust; (3) small listcos in the eastern region have the highest credit qualifi-cation compared with those in other geographical regions, and non-backdoor-listed small listcos have the highest credit qualification com-pared with those listed in other ways.

Key words: average accuracy, small listcos, default prediction, indicator selection

董冰洁, 周颖, 李继哲, 王珊珊. 基于平均准确度和贝叶斯准则指标遴选的企业违约预测研究——以中国上市小企业为例[J]. 管理评论, 2024, 36(5): 12-24.

Dong Bingjie, Zhou Ying, Li Jizhe, Wang Shanshan. An Enterprise Default Prediction Model Based on the Indicators Selected by Average Accuracy and Bayesian Information Criterion-Taking China's Small Listed Companies as an Example[J]. Management Review, 2024, 36(5): 12-24.

参考文献

[1] 白雪鹏,赵志冲.似然函数视角下小企业信用风险最优评价指标体系的建立[J].运筹与管理, 2023,32(4):155-161 Bai X. P., Zhao Z. C. Optimal Credit Risk Evalution Index System of Small Business from the Perspective of Likelihood Fuction[J]. Operations Research and Management, 2023,32(4):155-161
[2] Zhang Z. P., Chi G. T., Colombage S., et al. Credit Scoring Model Based on a Novel Group Feature Selection Method:The Case of Chinese Small-sized Manufacturing Enterprises[J]. Journal of the Operational Research Society, 2022,73(1):122-138
[3] 冯盼峰,温永仙.基于随机森林算法的两阶段变量选择研究[J].系统科学与数学, 2018,38(1):119-130 Feng P. F., Wen Y. X. Two-stage Stepwise Variable Selection Based on Random Forests[J]. Systems Science and Mathematics, 2018,38(1):119-130
[4] 曹桃云.基于随机森林的变量重要性研究[J].统计与决策, 2022,38(4):60-63 Cao T. Y. Study on the Importance of Variables Based on Random Forest[J]. Statistics and Decision Making, 2022,38(4):60-63
[5] 毕文杰,扶春娟.基于机器学习的Airbnb房源价格预测及影响因素研究——以北京市为例[J].运筹与管理, 2022,31(9):217-224 Bi W. J., Fu C. J. Price Estimation and Determinants Research of Airbnb with Machine Learning:Based on Data from Beijing[J]. Operations Research and Management, 2022,31(9):217-224
[6] Speiser J. L., Miller M. E., Tooze J., et al. A Comparison of Random Forest Variable Selection Methods for Classification Predic-tion Modeling[J]. Expert Systems with Applications, 2019,134:93-101
[7] 迟国泰,章彤,张志鹏.基于非平衡数据处理的上市公司ST预警混合模型[J].管理评论, 2020,32(3):3-20 Chi G. T., Zhang T., Zhang Z. P. Special Treatment Waring Hybrid Model Dealing with Imbalanced Data of Chinese List Compa-nies[J]. Management Review, 2020,32(3):3-20
[8] 赵志冲,严丽霞.多维数据下小企业违约风险过程性评价研究[J].中国管理科学, 2023,31(5):84-92 Zhao Z. C., Yan L. X. Research on Process Evaluation of Default Risk of Small Enterprises under Multidimensional Data[J]. Chi-nese Management Science, 2023,31(5):84-92
[9] Liang D., Tsai C., Lu H. R., et al. Combining Corporate Governance Indicators with Stacking Ensembles for Financial Distress Prediction[J]. Journal of Business Research, 2020,120:137-146
[10] Stevenson M., Mues C., Bravo C. The Value of Text for Small Business Default Prediction:A Deep Learning Approach[J]. Eu-
[11] Tsai C., Sue K., Hu Y., et al. Combining Feature Selection, Instance Selection, and Ensemble Classification Techniques for Im-proved Financial Distress Prediction[J]. Journal of Business Research, 2021,130:200-209
[12] 潘明道,周颖,迟国泰,等.基于Fisher判别的小型工业企业债信评级模型及实证[J].管理评论, 2018,30(3):15-28 Pan M. D., Zhou Y., Chi G. T., et al. Small Industrial Enterprises' Credit Rating Model and Empirical Analysis Based on Fisher Discriminant[J]. Management Review, 2018,30(3):15-28
[13] 于善丽,迟国泰,姜欣.基于指标体系违约鉴别能力最大的小企业债信评级体系及实证[J].中国管理科学, 2020,28(6):38-50 Yu S. L., Chi G. T., Jiang X. Small Enterprise Facility Rating on the Maximum Discrimination of Indicator System[J]. Chinese Management Science, 2020,28(6):38-50
[14] 吴辰文,梁靖涵,王伟,等.基于递归特征消除方法的随机森林算法[J].统计与决策, 2017,(21):60-63 Wu C. W., Liang J. H., Wang W., et al. Random Forest Algorithm Based on Recursive Feature Elimination[J]. Statistics and Decision Making, 2017,(21):60-63
[15] Yan X., Jia M. Intelligent Fault Diagnosis of Rotating Machinery Using Improved Multiscale Dispersion Entropy and Mrmr Feature Selection[J]. Knowledge-Based Systems, 2019,163:450-471
[16] Shukla A. K., Singh P., Vardhan M. Gene Selection for Cancer Types Classification Using Novel Hybrid Metaheuristics Approach[J]. Swarm and Evolutionary Computation, 2020,54:100661
[17] Duffie D., Singleton K. J. Modeling Term Structures of Defaultable Bonds[J]. Review of Financial Studies, 1999,12(4):687-720
[18] Altman E. I., Brady B., Resti A., at el. The Link between Default and Recovery Rates:Theory, Empirical Evidence, and Impli-cations[J]. Journal of Business, 2005,78(6):2203-2227
[19] Sun J., Li J., Fujita H. Multi-class Imbalanced Enterprise Credit Evaluation Based on Asymmetric Bagging Combined with Light Gradient Boosting Machine[J]. Applied Soft Computing, 2022,130:109637
[20] Wang L. Imbalanced Credit Risk Prediction Based on SMOTE and Multi-kernel FCM Improved by Particle Swarm Optimization[J]. Applied Soft Computing, 2022,130:108153
[21] Veganzones D., Severin E., Chlibi S. Influence of Earnings Management on Forecasting Corporate Failure[J]. International Jour-nal of Forecasting, 2023,39(1):123-143
[22] Shen F., Zhao X., Kou G., at al. A New Deep Learning Ensemble Credit Risk Evaluation Model with an Improved Synthetic Mi-nority Oversampling Technique[J]. Applied Soft Computing, 2021,98:106852
[23] Elhoseny M., Metawa N., Sztano G., et al. Deep Learning-based Model for Financial Distress Prediction[J]. Annals of Opera-tions Research, 2022,58(5):102673
[24] 陈艺云.基于信息披露文本的上市公司财务困境预测:以中文年报管理层讨论与分析为样本的研究[J].中国管理科学, 2019,27(7):23-34 Chen Y. Y. Forecasting Financial Distress of Listed Companies with Texual Content of the Information Disclosure:A Study Based MD&A in Chinese Annual Reports[J]. Chinese Management Science, 2019,27(7):23-34
[25] 王玉龙,周榴,张涤霏.企业债务违约风险预测——基于机器学习的视角[J].财政科学, 2022,(6):62-74 Wang Y. L., Zhou L., Zhang D. F. Enterprise Debt Default Risk Prediction-Based on the Perspective of Machine Learning[J]. Financial Science, 2022,(6):62-74
[26] Bai Z., Choi K. P., Fujikoshi Y. Consistency of AIC and BIC in Estimating the Number of Significant Components in High-dimensional Principal Component Analysis[J]. Annals of Statistics, 2018,46(3):1050-1076
[27] Ko P., Lin P., Do H., et al. P2P Lending Default Prediction Based on AI and Statistical Models[J]. Entropy, 2022,24(6):801
[28] 董冰洁,迟国泰.基于情感数据的违约判别研究[J].中国管理科学, 2023,31(4):111-120 Dong B. J., Chi G. T. Study on Default Prediction Based on Sentiment Date[J]. Chinese Management Science, 2023,31(4):111-120
[29] 迟国泰,董冰洁.基于借款描述的违约判别研究[J].管理评论, 2022,34(11):261-271 Chi G. T., Dong B. J. Research on Default Prediction Based on Loan Description[J]. Management Review, 2022,34(11):261-271
[30] 迟国泰,董冰洁.基于混合模型的利润驱动违约判别临界点研究[J].运筹与管理, 2022,31(9):196-201 Chi G. T., Dong B. J. Research on Cut-off Point of Profit-driven Default Judgment Based on Mixed Model[J]. Operations Re-search and Management, 2022,31(9):196-201
[31] Ding Y., Song X., Zen Y. Forecasting Financial Condition of Chinese Listed Companies Based on Support Vector Machine[J]. Expert Systems with Applications, 2008,34(4):3081-3089
[32] Du X., Li W., Ruan S., et al. Cus-heterogeneous Ensemble-Based Financial Distress Prediction for Imbalanced Dataset with En-semble Feature Selection[J]. Applied Soft Computing, 2020,97:106758
[33] 林宇,吴庆贺,李昊,等.基于Twin-SVR的公司违约风险预测研究[J].管理评论, 2019,31(11):33-43 Lin Y., Wu Q. H., Li H., et al. Research on Corporate Default Risk Prediction Based on Twin-SVR[J]. Management Review, 2019,31(11):33-43
[34] 王超.深度学习在行业指数技术分析中的应用研究[J].管理评论, 2021,33(3):75-83 Wang C. Study on Application of Deep Learning in Technical Analysis of Sector Indexes[J]. Management Review, 2021,33(3):75-83
[35] Chang Y., Chang K., Wu G. Application of Extreme Gradient Boosting Trees in the Construction of Credit Risk Assessment Mod-els for Financial Institutions[J]. Applied Soft Computing, 2018,73:914-920 ropean Journal of Operational Research, 2021,295(2):758-771