The　traditional　ensemble　modeling　method　has　problems　such　as　complex　calculation,　low　precision,　and　insufficient　generalization　performance.　Therefore,　this　article　proposes　three　modeling　methods　based　on　Bayesian　inference　and　binary　tree　fusion　ensembled　(BTFE).　Among　them,　Bayesian　inference　are　used　as　the　selection　rule　and　to　distinguish　the　adaptability　of　sub-model.　Three　different　modeling　methods　based　on　Bagging　and　binary　tree　are　proposed　to　replace　the　traditional　complex　ensemble　model,　which　reduces　the　training　difficulty　of　the　model　and　improves　the　interpretability　of　the　ensemble　model.　The　binary　tree　and　Bayesian　inference　fusion　method　is　proposed　to　construct　the　selective　ensemble　model.　Experimental　results　on　benchmark　data　sets　show　the　effectiveness　of　the　proposed　method.　©　2023　IEEE.