Abstract In this paper, we treat the problem of credit default risk or risk of non-repayment in banks using credit scoring models. As the methods currently used have some gaps in predicting the solvency of loan applicants, which could cause in losses for the banks, our contribution is to propose a new credit scoring method based on Machine Learning algorithms. We adopt two strategies: first, we hybridize Genetic Algorithms (GA) with Support Vector Machine (SVM) and Multi-Layer Perceptron (MLP) to evaluate impact of GA on prediction performance; and second, we test SVM and MLP with their hyperparameters, and then hybridize MLP with Artificial Neural Network (ANN). To compare our method with the methods proposed in [28], we realized simulations using Python on our banking dataset. The generated results show that the hybridization with GA yields less significant results compared to the strategy of SVM, MLP with their hyperparameters, and MLP-ANN that generate the improved values of AUC, Accuracy, confusion matrix and F1-Score compared to [28]. Furthermore, even for our database the same metrics is also significant with best values.
Alan : Mühendislik
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|