The　extended　belief　rule　base　(EBRB)　model　has　emerged　as　one　of　the　most　rapidly　evolving　branches　in　the　domain　of　artificial　intelligence　in　recent　years.　Nevertheless,　it　is　a　great　challenge　that　how　to　automatically　divide　EBRB　into　multiple　EBRBs　with　the　motivation　to　reduce　the　modelling　complexity　of　EBRB　models.　Hence,　this　research　presents　an　adaptive　modelling　method　for　constructing　a　novel　EBRB　model,　called　Ada-EBRB　model.　The　core　idea　of　the　proposed　adaptive　modelling　method　is　to　explore　the　latent　connections　among　rules　based　on　the　density　peak　clustering　(DPC)　algorithm,　which　can　achieve　the　automatic　determination　of　the　number　of　EBRBs　in　the　Ada-EBRB　model.　Moreover,　the　proposed　modelling　method　provides　a　parallel　solution　to　optimise　the　parameters　of　all　EBRBs.　The　experimental　results　demonstrate　that　the　Ada-EBRB　model　has　both　excellent　modelling　complexity　and　inference　accuracy　better　than　the　existing　models.　©　2025　Northeastern　University,　China.