Table 3.

Mean and standard deviation of area under the receiving operator curve (AUC) values obtained for binary mental workload classification when considering a model trained with data from all subjects, one model per subject and leave-one-subject-out validation.

Modality	Condition	AUC—Mixed subjects	AUC—Intra-subject	AUC—Leave-one-subject-out
EEG	No	0.774 ± 0.008	0.823 ± 0.139	0.523 ± 0.073
	Med	0.936 ± 0.004	0.927 ± 0.110	0.511 ± 0.093
	High	0.945 ± 0.004	0.929 ± 0.099	0.518 ± 0.112
	All	0.868 ± 0.004	0.805 ± 0.147	0.500 ± 0.049
Temperature	No	0.679 ± 0.026	0.846 ± 0.258	0.514 ± 0.142
	Med	0.641 ± 0.028	0.830 ± 0.279	0.509 ± 0.125
	High	0.656 ± 0.026	0.787 ± 0.303	0.506 ± 0.122
	All	0.594 ± 0.016	0.632 ± 0.282	0.514 ± 0.069
GSR	No	0.712 ± 0.025	0.882 ± 0.233	0.498 ± 0.144
	Med	0.761 ± 0.027	0.923 ± 0.169	0.522 ± 0.159
	High	0.692 ± 0.026	0.827 ± 0.256	0.557 ± 0.135
	All	0.661 ± 0.015	0.711 ± 0.264	0.519 ± 0.068
BVP	No	0.580 ± 0.029	0.720 ± 0.255	0.512 ± 0.109
	Med	0.624 ± 0.029	0.751 ± 0.258	0.539 ± 0.078
	High	0.584 ± 0.028	0.744 ± 0.249	0.494 ± 0.098
	All	0.562 ± 0.016	0.644 ± 0.183	0.481 ± 0.065
ECG	No	0.778 ± 0.016	-	-
	Med	0.780 ± 0.018
	High	0.753 ± 0.026
	All	0.748 ± 0.011
Breathing	No	0.913 ± 0.011	-	-
	Med	0.892 ± 0.013
	High	0.903 ± 0.012
	All	0.865 ± 0.011