Modeling with logistic regression for credit grant analysis

• Main Author: Beserra, Rafaella Santos
• Publication Date: 2022
• Other Authors: Barbosa, Nyedja Fialho Morais, Peixoto, Ana Patricia Bastos, Morais Xavier, Érika Fialho, Jale, Jader Silva, Xavier Júnior, Sílvio Fernando Alves
• Format: Article
• Language: por
• Source: Research, Society and Development
• Download full: https://rsdjournal.org/index.php/rsd/article/view/29761
• topic: Minería de datos; Curva ROC; Probabilidad.; Data mining; ROC curve; Probability.; Mineração de dados; Curva ROC; Probabilidade.

With the advancement of Big Data and the growing number of large masses of data in the most diverse areas of study, data mining techniques become increasingly necessary to obtain accurate and robust statistical information. This study aimed to show the efficiency of logistic regression as a data mining technique in obtaining a useful and statistically effective model in the analysis of customers for granting bank credit. The data comes from the Machine Learning Repository’s at the University of California-Irvin UCI. The database was divided into two groups: training and testing. The adjusted model was selected using the stepwise method in the R program. The model met the expectations of goodness of fit, with an accuracy of approximately 72% in discriminating non-defaulting from non-defaulting customers, sensitivity of 87% of the 140 non-defaulting customers, the model was correct 122 and specificity of 38%. The ROC curve had an area of 0.847, suggesting an effective fit.