Table 2.

Four-class classification performances of NB, LDA, and SVM when trained with fNIRS and behavioral features (i.e., GE_cm, GE_dm, and CQ). Each performance metric is represented in percentages (%) as the mean value across all runs ± standard deviation of the mean. Bold-typed results denote significantly higher performance of the corresponding algorithm with respect to the results when the algorithm is fed with fNIRS only features.

Method	Accuracy	Precision	Recall	Specificity	F1-Score
NB	84.68 ± 1.3	85 ± 0.01	83 ± 0.01	95 ± 0.01	84 ± 0.01
LDA	83.8 ± 1.6	84 ± 1.1	83 ± 1.4	94 ± 0.04	84 ± 1.2
SVM	85 ± 1.77	86 ± 1.6	84 ± 1.7	95 ± 0.5	85 ± 1.7