TPOT on Increasing the Performance of Credit Card Application Approval Classification
Aji Gautama Putrada, Etika Khusnul Laeli, Syafrial Fachri Pane, Nur Alamsyah, Mohamad Nurkamal Fauzan · 2022
Credit approval or credit card approval with the help of machine learning classification has become a hot topic in recent years. However, credit card approval classifications from previous papers have limited performance. Therefore, we propose evaluating tree-based pipeline optimization (TPOT) as the classification model creation automation for credit card approval in this study. The data used is 690, namely the credit card approval dataset sourced from Kaggle, which we obtained from the University of California learning machine and is publicly available. Then we compare the TPOT result model with the credit card approval classification from the related paper as a benchmark, namely naïve Bayes, support vector machine (SVM), and decision tree. The test results show that the TPOT model of credit card approval classification has higher accuracy than the benchmark model of the related research. The results of the TPOT have accuracy = 0.89. While the accuracy of naïve Bayes, SVM, and decision tree are 0.823, 0.852, and 0.796, respectively. Then we also see that the TPOT model results can produce a receiver operating curve (ROC) with the highest area under curve (AUC) value compared to the benchmark method, which is 0.940. AUC naive Bayes method, SVM, and decision tree are 0.897, 0.918, and 0.796, respectively.