Assessment and early warning of personal credit default behavior: Analysis based on logistic regression model
Yusheng Kang
School of Mathematics and Statistics, Henan University of Technology
Xianghui Li
School of Mathematics and Statistics, Henan University of Technology
Hanyan Zhang
School of Mathematics and Statistics, Henan University of Technology
DOI: https://doi.org/10.59429/bam.v8i2.14519
Keywords: credit default; information gain; logistic regression; demographic characteristics
Abstract
With the rapid development of digital finance, the assessment and early warning of personal credit default behavior have become critical components of financial institutions' risk control. Based on a real-world credit card transaction dataset containing over 65,000 records, this study employs the information gain method to select key predictive indicators and applies a logistic regression model to evaluate and provide early warnings for credit default behavior.The findings are as follows: First, the core factors influencing personal credit default behavior are primarily concentrated on borrowers' historical credit records and debt status, rather than on demographic characteristics. Second, the logistic regression early warning model constructed in this paper achieves a sound balance between predictive accuracy and risk identification capability, successfully identifying all defaulting clients. This result indicates that the model can effectively control the type of misclassification that carries the highest risk cost in financial risk management—Namely, misclassifying a "bad" client as a "good" one—Thereby demonstrating high practical value.Therefore, the personal credit risk early warning model based on logistic regression and the information gain method can accurately identify credit default behaviors among high-risk borrowers while maintaining interpretability. It is recommended that financial institutions, in their risk control practices, place particular emphasis on borrowers' historical credit blemishes, large outstanding debts, and records of bounced checks, supplemented by influencing factors such as age and household size.
References
[1] Fisher, R. A. (1936). The use of multiple measurements in taxonomic problems. Annals of Eugenics, 7, 179-188.
[2] Durand, H. F., & Soons, S. G. (1941). Repudiation of family trust settlements. Columbia Law Review, 41(2).
[3] Orgler, Y. E. (1970). Practical financial management models: Reply. Management Science, 16(12).
[4] Wiginton, J. C. (1980). A note on the comparison of logit and discriminant models of consumer credit behavior. Journal of Financial and Quantitative Analysis, 15(3), 757-770.
[5] Xu, S. F., & Wang, Y. C. (2003). Logistic model in personal credit evaluation. Journal of Tianjin Institute of Light Industry, (S1), 46-49. (in Chinese)
[6] Chen, X. L. (2020). Research on credit default risk assessment model of P2P borrowers based on Logistic regression model [Master's thesis]. Shanghai International Studies University. (in Chinese)
[7] Wang, M. J. (2015). Credit risk assessment of P2P lending platform borrowers based on Logistic regression model [Master's thesis]. Beijing Foreign Studies University. (in Chinese)
[8] Yu, G. L. (2017). Research on credit risk assessment of P2P online lending platform borrowers based on Logistic regression model [Master's thesis]. Lanzhou University. (in Chinese)
[9] Zhang, L. K. (2016). Research on personal credit risk assessment based on P2P online lending platform [Master's thesis]. Harbin University of Science and Technology. (in Chinese)
[10] Lin, C. D., & Peng, G. L. (2007). Application of random forest in determining the index system for corporate credit evaluation. Journal of Xiamen University (Natural Science Edition), 46(2), 199-203. (in Chinese)