Table 4.

Pooled testing efficiency prediction results (number of tests divided by the number of subjects tested)

	Training Datasets		Production Dataset
	Accuracy Optimized	Efficiency Optimized	Accuracy Optimized	Efficiency Optimized
Spring dataset training (10% negatives; all features)
Logistic regression	0.995	0.766	0.656	0.636
Support vector machine	1.000	0.757	1.000	0.631
Summer dataset training (40% negatives, features without age)
Logistic regression	0.970	0.779	0.720	0.689
Support vector machine	0.999	0.793	0.999	0.668
Spring+Summer dataset training (positives subset to reach 5%, top features)
Logistic regression	0.957	0.807	0.973	0.643