Research on Default Prediction for Credit Card Users Based on XGBoost-LSTM Model

Table 3

Model evaluation indicators of XGBoost algorithm models using different features.

Features	Train AUC	Train ACC	Test AUC	Test ACC	Test recall	Test precision
Without transaction features	0.716	0.801	0.710	0.799	0.008	0.493
The first stage	0.764	0.821	0.759	0.819	0.163	0.714
The first and second stages	0.895	0.873	0.893	0.870	0.507	0.768