. 2021 Feb 5;13:623607. doi: 10.3389/fnagi.2021.623607

Table 2.

Validation and Test results of the Model B0 with different feature sets on the classification task.

Feature set	Val/Test	Class	Recall	Precision	F1 score	Accuracy
eGeMAPS	5-fold CV	Non-AD	0.527 ± 0.120	0.710 ± 0.151	0.581 ± 0.058	0.635 ± 0.034
	5-fold CV	AD	0.745 ± 0.156	0.618 ± 0.029	0.667 ± 0.058	0.635 ± 0.034
	Test set	Non-AD	0.7500	0.5625	0.6428	0.5833
	Test set	AD	0.4166	0.6250	0.5	0.5833
emobase	5-fold CV	Non-AD	0.659 ± 0.094	0.704 ± 0.168	0.663 ± 0.057	0.665 ± 0.082
	5-fold CV	AD	0.673 ± 0.219	0.664 ± 0.049	0.652 ± 0.125	0.665 ± 0.082
	Test set	Non-AD	0.6667	0.6400	0.6530	0.6458
	Test set	AD	0.6250	0.6521	0.6382	0.6458
ComParE	5-fold CV	Non-AD	0.441 ± 0.176	0.534 ± 0.139	0.475 ± 0.148	0.533 ± 0.129
	5-fold CV	AD	0.625 ± 0.144	0.538 ± 0.132	0.573 ± 0.124	0.533 ± 0.129
	Test set	Non-AD	0.5833	0.5185	0.5490	0.5208
	Test set	AD	0.4583	0.5238	0.4888	0.5208

Bold values represent the validation and test accuracies of best performing model amongst the models under consideration in the respective table.