Table 2.

Classification analysis results of individual features where supervised ML was performed using logistic regression from the Classifier App while unsupervised ML was performed using k-means; the results for the supervised ML show the mean of all algorithms that were trained/tested. A – accuracy (%), 95% CI – confidence interval (%) SE – sensitivity (%), SP – specificity (%), AUC – area under the ROC curve.

Features	t-test (p-score)	Supervised ML					Unsupervised ML
		Training + Testing
		AUC (Training only)	A	CI	SE	SP	A	CI	SE	SP
	0.001	0.85	85.7	15.0	84.6	87.5	76.2	18.2	88.9	66.7
	0.0099	0.86	67.7	20.2	72.7	60.0	76.2	18.2	100	64.3
	0.3146	0.61	33.3	20.2	42.8	0.0	57.1	21.2	71.4	50
	0.0055	0.93	85.7	15.0	90.9	80.0	76.2	18.2	100	64.3
	0.0244	0.86	66.7	20.2	66.7	66.7	66.7	21.2	85.7	57.1
	0.3489	0.24	52.4	21.4	55.6	33.3	47.6	21.4	60	45.4
	0.0006	0.86	81.0	16.8	78.6	85.7	85	15	84.6	87.5
	10−6	0.89	90.5	12.5	91.7	88.9	95.2	9.1	92.3	100
	0.0104	0.8	81.0	16.8	78.6	86.7	80.9	16.8	78.6	85.7
	0.0006	0.84	85.7	15.0	84.6	87.5	90.5	12.5	91.7	88.9
	0.0006	0.89	81.0	16.8	83.3	77.8	85.7	15	80	100
	0.0011	0.85	76.2	18.2	81.8	70.0	80.9	16.8	90	72.7
SBP	0.3327	0.4	47.6	21.4	52.9	75.0	52.4	21.4	50	55.5
DBP	0.8195	0.57	52.4	21.4	55.0	0.0	57.1	21.2	33.3	75