Multinomial Logistic Regression and Spline Regression for Credit Risk Modelling

Research output: Contribution to journalConference articlepeer-review

5 Citations (Scopus)


Regression modelling has been adapted in retail banking because of its capability to analyze the continuous and discrete data. It is an important tool for credit risk scoring, stress testing and credit asset evaluation. In this paper, the approach used is multinomial logistic regression model to gain the information regarding the factors that affect the occurrence of default and attrition events on credits. In addition, this paper will also introduce spline regression approach using truncated power basis to model the hazard functions of default and attrition events. The flexibility of spline function allows us to model the nonlinear and irregular shapes of the hazard functions. Then, by using spline regression and multinomial logistic regression model, there will be a better result and interpretation. There are several advantages by using those both models. First, by using the flexible spline regression function, it can model nonlinear and irregular shapes of the hazard functions. Second, it is easy to understand and implement, and its simple parametric form from multinomial logistic regression model can make it easy in model interpretation. Third, the multinomial logistic regression model has the ability to do prediction. Furthermore, by using a credit card dataset, we will demonstrate how to build these models, and we also provide statistical explanatory and the prediction accuracy of multinomial logistic regression model in classifying customers based on the prediction of default and attrition is 95.3%.

Original languageEnglish
Article number012019
JournalJournal of Physics: Conference Series
Issue number1
Publication statusPublished - 4 Dec 2018
Event2nd Mathematics, Informatics, Science and Education International Conference, MISEIC 2018 - Surabaya, Indonesia
Duration: 21 Jul 2018 → …


Dive into the research topics of 'Multinomial Logistic Regression and Spline Regression for Credit Risk Modelling'. Together they form a unique fingerprint.

Cite this