A Classifier Evaluation for Payments’ Default Predictions in a Brazilian Retail Company
This article presents an investigation about the performance of classification algorithms used for predicting payments’ default. Classifiers used for modelling the data set include: Logistic Regression; Naive-Bayes; Decision Trees; Support Vector Machine; k-Nearest Neighbors; Random Forests; and Artificial Neural Networks. These classifiers were applied to both balanced and original data using the Weka data mining tool. Results from experiments revealed that Logistics Regression and Naive Bayes classifiers had the best performance for the chosen data set.
KeywordsData mining Classifier algorithms Area under curve Logistic regression
The authors would like to thank: (1) the Brazilian Aeronautics Institute of Technology (ITA); (2) the Casimiro Montenegro Filho Foundation (FCMF); the Software Engineering Research Group (GPES) members; and the 2RP Net Enterprise for their infrastructure, data set, assistance, advice, and financial support for this work.
- 1.V. García, A.I. Marqués, J.S. Sánchez, An insight into the experimental design for credit risk and corporate bankruptcy prediction systems. J. Intell. Inf. Syst. 44(1), 159–189 (2015)Google Scholar
- 2.J. Abellán, J.G. Castellano, A comparative study on base classifiers in ensemble methods for credit scoring. Expert. Syst. Appl. 73, 1–10 (2017)Google Scholar
- 3.K. Kennedy, Credit scoring using machine learning, Doctoral thesis, Dublin Institute of Technology, 2013Google Scholar
- 5.M.C. Aniceto, Estudo comparativo entre técnicas de aprendizado de máquina para estimação de risco de crédito, Dissertação (Mestrado em Administração), Universidade de Brasília, Brasília, Brazil, 2016Google Scholar