A　new　method,　based　on　belief　rule　base(BRB)　was　proposed　for　constructing　a　classification　system　with　high　performance　for　classification　problems.　The　existing　belief　rule　base　classification　system　(BRBCS)　is　flawed　because　its　classification　accuracy　is　limited　to　the　partition　number,　the　parameter　training　method　needs　the　number　of　rules　given　in　advance,　and　the　reasoning　process　does　not　reflect　the　correlation　between　characteristics　and　results.　The　belief　rule　base　inference　method　was　thus　proposed　for　classification　based　on　the　differential　evolutionary　algorithm　(DEBRM)　to　solve　the　classification　problems.　The　proposed　method　consists　of　two　procedures:　belief　rule　base　classification　system　(BRBCS)　and　parameter　training　method.　The　new　method　first　introduced　the　construction　strategy　of　BRBCS　to　determine　the　number　of　rules.　Then,　belief　reasoning　method　was　adopted　as　the　inference　engine.　Finally,　the　training　model　for　classification　which　is　combined　with　differential　evolutionary　algorithm　was　built.　In　the　experiment　analysis,　the　effectiveness　of　the　method　was　validated　by　comparing　it　with　the　existing　parameter　training　method,　and　the　rationality　of　parameters　training　in　comparison　of　other　belief　rule　base　methods　for　different　number　of　partitions.　The　classification　results　show　that　the　proposed　method　is　effective　and　reasonable.