Depressive　disorders　are　heterogeneous　in　symptoms　and　course　of　disease,　while　it　is　difficult　to　reveal　the　neurophysiological　subtypes.　In　this　study,　we　combined　EEG　data　and　machine　learning　methods　to　identify　subtypes　of　depression,　and　further　evaluated　the　rationality　and　reliability　of　subtypes.　The　results　showed　that　the　left-right　asymmetry　of　prefrontal　lobe　in　alpha　band　clustered　32　patients　with　depression　into　three　different　subtypes,　each　type　showed　unique　clinical　characteristics.　High　classification　effects　could　be　obtained　through　decision　tree　and　logistic　regression　model.　The　study　further　investigated　the　new　way　to　identify　neurophysiological　subtypes　of　depression　with　machine　learning　methods.　This　study　also　shows　the　important　value　of　the　classic　EEG　index　of　depression,　namely　the　left-right　asymmetry　of　prefrontal　lobe　in　alpha　band,　in　the　identification　of　depression　subtypes.