Academic　prediction　is　an　important　management　means　to　strengthen　the　construction　of　study　style　and　improve　the　quality　of　talent　training　in　Colleges　and　universities.　In　view　of　the　current　colleges　and　universities　lack　of　scientific　and　effective　methods　in　the　undergraduate　academic　prediction,　based　on　deep　learning　of　undergraduate　academic　prediction　model,　the　model　mainly　includes　the　construction　of　feature　reconstruction　and　prediction　model　of　the　two　modules,　the　feature　extraction　module,　including　the　selection　of　students,　characteristics　of　format　conversion　and　dimension　recombination;　model　the　convolutional　neural　network　the　characteristics　of　the　input　after　the　reorganization　of　the　improved　training　and　validation,　the　model　parameters　were　optimized,　and　then　use　the　model　to　forecast.　Taking　the　data　from　2007　to　2010　students　in　a　university　in　Beijing　as　training　set,　the　model　is　trained,　and　2011　level　students　are　selected　as　the　validation　set　to　optimize　the　model　parameters,　and　the　2012　level　student　data　set　is　used　to　evaluate　the　model.　The　results　show　that　the　undergraduate　academic　prediction　model　based　on　CNN　neural　network　has　higher　prediction　accuracy.　And　then　help　colleges　and　universities　to　optimize　the　teaching　management　methods　and　improve　the　quality　of　teaching.　©　2018　IEEE.