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

Table 6

Performance of hybrid CNN-RF with learning scheme = 80 : 20.

Proposed methods	Criterion	Accuracy	Precision	Recall	F1-score	AUC
CNN–RF (max-depth = 5)	Gini	92.17	90.16	93.38	91.74	92.10
CNN–RF (max-depth = 5)	Entropy	91.00	89.64	91.53	90.58	90.95
CNN–RF (max-depth = 4)	Gini	87.58	81.17	92.16	86.32	87.37
CNN–RF (max-depth = 4)	Entropy	87.92	81.87	92.22	86.74	95.19

Note. n-Estimators = 80, different criterion, and different max-depth.