. 2024 Sep 13;3:e48588. doi: 10.2196/48588

Table 3.

Prediction model performance on joint data with cross-validation. For completeness, all models are included.

Candidate machine learning model			Specificity, %		Sensitivity, %		Precision or PPV^a, %		NPV^b, %		F₁-score		Balanced accuracy, %		Cost (US $)
Random forest model weights
	1	100		4.7		—^c		99.5		—		52.35		659,016
	200	77.2		78.2		1.77		99.85		0.0346		77.75^d		203,659.2
	300	77.1		75.9		1.7		99.83		0.0332		76.5		220,284
	500	79.1		74.1		1.82		99.83		0.0355^d		76.61		227,836.8
	1000	79.2		71.7		1.77		99.81		0.0345		75.51		243,777.6
Decision tree model weights
	1	99.3		10		—		99.53		—		54.66		623,973.6
	200	73.9		75.9		1.51		99.82		0.0296^d		74.87^d		227,908.8
	300	75		72.4		1.5		99.81		0.0294		73.66		249,696
	500	59.7		83.5		1.09		99.85		0.0215		71.62		208,080
	1000	68.6		74.1		1.22		99.8		0.0240		71.35		252,446.4
Logistic regression (L2) model weights
	1	100		0		—		99.48		—		50		691,560
	200	80.1		72.9		1.88		99.82		0.0366^d		76.53		233,596.8
	300	71.6		83.5		1.51		99.88		0.0296		77.6^d		180,151.2
	500	58.1		90.6		1.12		99.91		0.0221		74.36		162,964.8
	1000	37.1		93.5		0.77		99.91		0.0152		65.33		191,757.6
Logistic regression (L1) model weights
	1	100		0		—		99.48		—		50		691,560
	200	80.1		72.9		1.88		99.82		0.0366		76.52		233,625.6
	300	71.6		83.5		1.51		99.88		0.0296		77.59^d		180,151.2
	500	58.1		90.6		1.12		99.91		0.0221		74.36		162,964.8
	1000	45.7		88.2		0.84		99.87		0.0166		66.99		208,180.8
SVM^e
	1	99.7		7		—		99.51		—		53.39		643,384.8
	200	78.7		72.9		1.76		99.82		0.0343^d		75.84		236,800.8
	300	71.1		84.1		1.5		99.88		0.0294		77.61^d		177,393.6
	500	60.2		88.2		1.14		99.89		0.0225		74.21		174,456
	1000	50		92.9		0.87		99.92		0.0172		68.96		177,429.6

^aPPV: positive predictive value.

^bNPV: negative predictive value.

^cNot available.

^dBest model with respect to the specific metric.

^eSVM: support vector machines.