A novel EEG-based major depressive disorder detection framework with two-stage feature selection

Table 2 MDD detection performance of the proposed framework

Feature selection	Model	Sen	Prec	Spec	Acc	F₁
–	LR	0.4476	0.3776	0.6167	0.5557	0.4005
	SVM	0.4524	0.4809	0.7061	0.6139	0.4510
	LR-DF	0.5667	0.4728	0.6182	0.5984	0.5050
	SVM-DF	0.5333	0.5000	0.6894	0.6318	0.5144
PCC	LR	0.5619	0.5061	0.6530	0.6207	0.5235
	SVM	0.5381	0.5444	0.7591	0.6782	0.5306
	LR-DF	0.6238	0.6083	0.7394	0.6961	0.5988
	SVM-DF	0.5619	0.6442	0.7727	0.6969	0.5766
RFE	LR	0.7381	0.8571	0.9106	0.8466	0.7803
	SVM	0.9143	0.9667	0.9833	0.9579	0.9385
	LR-DF	0.8524	0.8298	0.8955	0.8800	0.8386
	SVM-DF	0.9429	1.0000	1.0000	0.9789	0.9667
PCC and RFE	LR	0.7381	0.8571	0.9106	0.8466	0.7803
	SVM	0.9714	0.9464	0.9652	0.9678	0.9581
	LR-DF	0.8238	0.9417	0.9636	0.9127	0.8713
	SVM-DF	0.9714	1.0000	1.0000	0.9895	0.9846

ISSN: 1472-6947