Credit-Risk Prediction Model Using Hybrid Deep—Machine-Learning Based Algorithms

Table 5

Performance of hybrid CNN–SVM (%) with different kernel tricks.

Proposed methods	Accuracy	Precision	Recall	F1-score	AUC
CNN–SVM (kernel = RBF)	96.50	96.72	95.68	96.19	96.52
CNN–SVM (kernel = linear)	93.60	97.60	89.40	93.32	93.91
CNN–SVM (kernel = polynomial)	87.40	83.19	88.60	85.81	87.07

Note. , gamma = 0.7, and learning scheme 80 : 20.