Table 6 Performance comparison in holdout test set.

Week	Model	F1-score		Recall		Accuracy	AUC
Week	Model	Approved	Reproved	Approved	Reproved	Accuracy	AUC
1	Multiple	0.63	0.49	0.56	0.60	0.57	0.63
1	Unique	0.73	0.51	0.73	0.50	0.65	0.66
3	Multiple	0.82	0.57	0.86	0.51	0.75	0.74
3	Unique	0.83	0.58	0.88	0.57	0.76	0.84
5	Multiple	0.82	0.55	0.78	0.63	0.74	0.76
5	Unique	0.92	0.74	0.94	0.68	0.86	0.92
7	Multiple	0.91	0.79	0.89	0.82	0.87	0.90
7	Unique	0.94	0.86	0.95	0.85	0.90	0.95
9	Multiple	0.90	0.74	0.95	0.65	0.85	0.90
9	Unique	0.97	0.94	0.96	0.93	0.95	1.00
11	Multiple	0.76	0.58	0.70	0.68	0.69	0.78
11	Unique	0.89	0.81	0.98	0.91	0.94	0.98
13	Multiple	0.92	0.67	0.92	0.64	0.87	0.88
13	Unique	0.98	0.96	1.00	0.93	0.99	0.97
15	Multiple	0.94	0.82	0.96	0.77	0.91	0.98
15	Unique	0.96	0.91	0.97	0.89	0.95	0.99

Unique model versus multiple models for each week. Significant values are in bold.

Quick links

Search