In　recent　years,　the　incidence　of　depression　is　increasing　year　by　year,　and　depression　lasts　too　long　after　the　onset　of　the　disease,　which　seriously　hinders　people＇s　normal　working　life.　In　this　study,　based　on　the　characteristics　of　scalp　EEG　signals,　we　compared　the　classification　performance　of　Support　Vector　Machine　(SVM),　Convolutional　Neural　Network　(CNN),　Convolutional　Neural　Network　(CNN),　and　Long　Short-Term　Memory　Network　(LSTM)　classification　models　for　depression　by　using　16　channels　of　clean　EEG　data,　and　the　accuracy　of　the　model　using　the　combination　of　CNN　and　LSTM　was　improved　about　9.21%,　which　confirms　that　the　use　of　LSTM　to　help　process　EEG　signals　and　improve　classification　is　real　and　effective,　and　the　effect　of　model　parameters　on　model　performance　is　discussed　at　the　end　of　the　paper　to　adjust　model　parameters　and　algorithms　to　improve　the　performance　of　classification.　©　2022　IEEE.