Since　classification　method　of　functional　magnetic　resonance　imaging　(fMRI)　data　can　not　effectively　extract　the　local　features,　the　classification　accuracy　is　seriously　affected.　To　solve　the　problem,　a　classification　model　of　fMRI　data　based　on　convolutional　neural　network　(CNN)　is　presented.　Firstly,　a　CNN　structure　is　designed,　and　a　restricted　boltzmann　machine　(RBM)　model　is　constructed　by　means　of　the　convolution　kernel　size.　Then,　the　interested　region　voxels　in　fMRI　data　are　employed　to　construct　and　form　input　data　to　pre-train　RBM,　and　the　relative　transformation　of　the　obtained　weight　matrix　is　executed　to　initialize　CNN　parameters.　Finally,　the　final　classification　model　is　obtained　by　training　the　whole　initialized　model.　The　results　on　Haxby　and　LPD　datasets　show　that　the　proposed　model　effectively　improves　the　classification　accuracy　of　fMRI　data.　©　2017,　Science　Press.　All　right　reserved.