EEG　state　classification　is　used　in　many　fields,　and　decision-making,　as　a　higher　cognitive　function　of　the　brain,　has　high　research　significance,　and　this　paper　is　mainly　to　study　the　state　of　decision-making.　Therefore,　we　study　the　classification　of　three　decision　states,　before-decision,　in-decision,　post-decision,　and　two　resting　states,　eye-opened,　eye-closed.　In　this　study,　three　methods　are　used　to　compare　the　classification　effects,　namely　DE+SVM,　DE+DGCNN,　EEGNet,　among　which　differential　entropy　(DE)　is　a　frequency　domain　feature,　which　can　extract　effective　features　in　EEG　emotion　recognition;　DGCNN　is　a　dynamic　graph　convolutional　neural　network　which　uses　DE　as　the　node　feature　and　dynamically　learns　the　adjacency　matrix　for　classification;　EEGNet　is　an　end-to-end　neural　network,　which　is　designed　to　be　used　in　multiple　experimental　paradigms.　The　above　3　methods　achieved　62.80%±9.67%,　78.70%±8.27%,　and　88.83±6.03%　accuracy　in　within-subject　classification　respectively.　Finally,　we　visualize　the　adjacency　matrix　learned　by　DGCNN　and　the　spatial　filter　learned　by　EEGNet　to　see　the　knowledge　learned　by　the　model.　©　2022　IEEE.