Table 4.

Sensitivity, specificity, and NPV by varying cut-off in Lasso, Random Forest and Gradient Boosting Machines.

LASSO
Cut-Off	Sensitivity	Specificity	NPV
0.05	0.96	0.29	0.98
0.10	0.91	0.51	0.98
0.15	0.84	0.68	0.97
0.20	0.76	0.80	0.96
0.50	0.29	0.98	0.91
Random Forest
Cut-Off	Sensitivity	Specificity	NPV
0.05	0.92	0.48	0.98
0.10	0.86	0.69	0.97
0.15	0.80	0.79	0.96
0.20	0.73	0.85	0.96
0.50	0.41	0.97	0.92
Gradient Boosting Machines
Cut-Off	Sensitivity	Specificity	NPV
0.05	0.90	0.59	0.98
0.10	0.81	0.77	0.97
0.15	0.74	0.85	0.96
0.20	0.67	0.89	0.95
0.50	0.41	0.97	0.92

Metrics with cut-off probabilities above 0.50 are not listed as the sensitivity drops to a very low level. NPV: Negative Predictive Value.