| dc.description.abstract | An essential method for systematically forecasting the unpredictable hazards in banking systems
is financial risk prediction. When clients recently pay the limit balance debt, credit card default
causes many banks to experience losses. An efficient risk prediction approach is required due to
the problems with poor timing and low accuracy in the current risk prediction methods. The
objective of this paper is to propose a risk prediction model of Extreme Gradient Boosting
(XGBoost), together with a comperative study of feature selection techniques to enhance the
accuracy of prediction. This paper also proposes a method of tuning hyperparameters to optimize
the prediction performance. Moreover, the results achieved are taken into comparation with
several machine learning algorithms such as Logistics Regression (LR), Decision Tree (DT), K
Nearest Neighbor (KNN). The proposed method have the accuracy of 77.43% and 67.36% of
AUC score, which shows that the proposed techniques have more advantages than other
mentioned techniques in predictive performance. | en_US |
