With　the　development　of　information　society,　tens　of　thousands　of　data　will　be　generated　at　any　time.　How　to　obtain　valuable　information　from　data　has　always　been　explored　by　scholars.　As　an　important　research　field　in　machine　learning,　data　classification　can　discover　classification　rules　and　make　data　prediction　by　analyzing　the　training　set　of　known　categories.　In　order　to　solve　the　problem　of　multi-classification　unbalanced　data,　we　use　DPC-based　data　preprocessing　method　to　optimize　the　classifier,　and　select　three　mainstream　classifiers,　namely　random　forest,　decision　tree,　k-nearest　neighbor　and　support　vector　machine　model　for　analysis　and　research.　It　is　worth　noting　that　in　the　KNN　model,　we　use　the　weighted　KNN　algorithm　based　on　kernel　function　for　prediction　training.　Specifically,　this　paper　first　uses　DPC　algorithm　to　quickly　select　the　cluster　center,　and　selects　the　optimal　clipping　threshold　through　experiments　to　pry　the　training　set,　uses　the　optimized　training　set　to　train　and　optimize　the　classifier,　and　finally　uses　the　classifier　to　predict　the　test　set.　In　order　to　better　test　the　model　accuracy,　we　choose　the　50%　cross　test.　Experimental　results　show　that　the　optimization　algorithm　can　improve　the　classifier　accuracy　in　99.9%　of　the　cases.　©　2023　IEEE.