With　the　rapid　development　of　machine　learning,　it　has　been　widely　used　in　many　industries　and　fields　to　solve　problems.　The　application　of　machine　learning　to　the　classification　of　network　traffic　is　also　a　new　research　direction.　However,　the　machine　learning　methods　used　in　the　past　cannot　classify　traffic　data　according　to　application　types.　In　this　paper,　a　method　of　random　forest　is　proposed　for　traffic　classification　and　identification.　As　a　new　idea　of　network　construction,　SDN　separates　the　data　plane　from　the　control　plane.　The　data　layer　collects　the　whole　network　state　information,　and　the　control　layer　realizes　the　centralized　control　of　the　network.　The　internal　structure　of　random　forest　is　actually　composed　of　many　decision　trees.　While　the　decision　tree　itself　has　been　well　represented　in　the　field　of　traffic　classification,　the　random　forest　model　further　improve　the　efficiency　of　classification　on　the　basis　of　the　decision　tree.　By　effectively　capturing　the　relevant　characteristics　of　the　original　data　set　and　training　the　random　forest　classification　model,　we　can　find　through　experiments　that　the　classification　of　network　traffic　using　random　forest　has　obvious　advantages　in　the　accuracy　of　classification　as　well　as　the　efficiency　and　stability　of　processing　large-scale　data　sets.　©　2018　IEEE.