TABLE 3.

Comparison with previous studies on attention recognition.

Authors	Attention task (levels)	Subjects	Methods	Window (seconds)	Brain regions (channels)	Validation	Accuracy (%)
							2-Levels
Chen et al. (2017)	Continuous performance task (high-attention, low-attention)	10	Temporal and entropy features—SVM	Trial length	Prefrontal (1)	3/4 train, 1/4 test	91.60
							3-Levels
Hu et al. (2018)	Randomly selected learning task (high, neutral, low)	10	Linear and nonlinear features—CFS+KNN	180	Central and temporal (6)	10 times 3-Fold CV	80.84
							2-Levels		3-Levels
Gaume et al. (2019)	Continuous performance task (easy, medium, and hard)	14	Power features—LDA	5	Whole brain (16)	Leave-one-subject-out	75		51.8
				30			85		64.8
							2-Levels (in-ear)		2-Levels (prefrontal)
Jeong and Jeong (2020)	Psychomotor vigilance tasks (attention, rest)	6	Temporal and spectral features—Echo State Network	0.5	In-ear (2) prefrontal (2)	Within-subject	81.16		82.44
						Cross-subject	64.00		65.70
						10-Fold CV	74.15		73.73
							2-Levels	3-Levels		4-Levels
Our study	AX-CPT (rest, LA, MA, HA)	42	Complexity—XGBoost	3	Frontal (13)	Leave-one-subject-out	76.30	70.49		64.69
						5-Fold CV	95.36	80.42		81.39

SVM, support vector machine; CFS, correlation-based feature selection; KNN, k-nearest-neighbor; LDA, Linear discriminant analysis; ESN, Echo State Network; CV, cross-validation.