Table 7.

Performance summary of classifying Low and High MWL with LgR, MLP, SVM and RF classifier models using EEG and MI-based feature on the holdout test set. In this task, the total number of observations was 1710, where low MWL was considered as the positive class. The number of observations with positive and negative class were 917 and 793, respectively. The highest accuracies obtained by using different feature sets are marked with (*).

Criteria	Using EEG-Based Features				Using MI-Based Features
	LgR	MLP	SVM	RF	LgR	MLP	SVM	RF
True Positive	736	776	342	864	688	715	576	783
False Negative	181	141	575	53	229	202	341	134
False Positive	362	293	138	157	410	230	146	175
True Negative	431	500	655	636	383	563	647	618
Sensitivity	0.80	0.85	0.37	0.94	0.75	0.78	0.63	0.85
Specificity	0.54	0.63	0.83	0.80	0.48	0.71	0.81	0.78
Precision	0.67	0.73	0.71	0.85	0.63	0.76	0.80	0.82
Recall	0.80	0.85	0.37	0.94	0.75	0.78	0.63	0.85
${\mathit{F}}_{\mathbf{1}}$ score	0.73	0.78	0.50	0.89	0.68	0.77	0.70	0.84
Accuracy	0.68	0.75	0.58	0.88 *	0.63	0.75	0.72	0.82 *
Balanced Accuracy	0.67	0.74	0.60	0.87 *	0.62	0.74	0.72	0.82 *