Global　warming　and　climate　change　are　gaining　traction　in　recent　years.　As　a　major　cause　of　global　warming,　carbon　emissions　were　centered　to　China＇s　climate　change　policy　initiatives.　Nevertheless,　the　existing　policy　discourse　has　yet　reached　a　consensus　on　the　optimal　modeling　method　for　carbon　emissions　prediction　that　is　well-informed　of　both　policy　goals　and　the　time-series　pattern　of　carbon　emissions.　This　paper　fills　the　gap　by　promoting　a　novel　data-driven　decision　model　for　carbon　emissions　prediction　that　is　based　on　the　extended　belief　rule　base　(EBRB)　inference　model.　The　new　decision　model　consists　of　three　components:　1)　an　indicator　integration　method,　which　aims　to　generate　a　few　group　indicators　from　a　large　number　of　statistical　indicators;　2)　a　new　EBRB　construction　method,　which　aims　to　consider　the　management　policy　goals　for　constructing　EBRB;　3)　a　new　ER-based　inference　method,　which　aims　to　predict　carbon　emissions　based　on　time　series　change　of　relevant　factors.　The　effectiveness　of　the　proposed　decision　model　has　been　tested　against　carbon　emissions　management　data　from　30　provinces　in　China.　Experimental　results　demonstrate　that　the　model　will　offer　powerful　reference　value　in　the　policy　decision-making　process,　which　will　help　to　meet　policy　requirements　for　carbon　emissions.　©　2022　Elsevier　Ltd