Table 7.

LOOCV performance for the transductive SVM (TSVM) model and logistic regression (LR) model

Model	Feature type	Contrast	Sensitivity (%)	Specificity (%)	Accuracy (%)	AUC	Pcorr	Scorr
TSVM	BoW44	Speech vs. Silence	88.9	28.6	62.5	0.80	0.36	0.68
		Noise vs. Silence	77.8	57.1	68.8	0.66	0.32	0.26
		Speech vs. Noise	33.3	71.4	50.0	0.55	0.03	0.09
		Combine	77.8	71.4	75.0	0.76	0.40	0.36
	BoW21	Speech vs. Silence	77.8	42.9	62.5	0.67	0.37	0.33
		Noise vs. Silence	77.8	57.1	68.8	0.74	0.31	0.27
		Speech vs. Noise	44.4	71.4	56.3	0.52	0.00	−0.18
		Combine	88.9	57.1	75.0	0.66	0.37	−0.02
LR	BoW44	Speech vs. Silence	44.4	71.4	56.3	0.51	0.08	−0.07
		Noise vs. Silence	77.8	42.9	62.5	0.62	0.27	0.17
		Speech vs. Noise	77.8	42.9	62.5	0.62	0.08	0.09
		Combine	33.3	71.4	50.0	0.56	−0.07	0.10
	BoW21	Speech vs. Silence	22.2	71.4	43.8	0.43	−0.09	−0.08
		Noise vs. Silence	33.3	71.4	50.0	0.51	−0.02	0.03
		Speech vs. Noise	55.6	42.9	50.0	0.54	−0.17	−0.21
		Combine	44.4	57.1	50.0	0.60	−0.07	0.05

LOOCV, Leave‐One‐Out Cross‐Validation; SVM, Support Vector Machine.