Rapid　traffic　development　brings　convenience　to　social　circulation,　but　the　number　of　fatalities　in　traffic　accidents　has　brought　great　pressure　on　traffic　safety　and　social　stability　management.　Therefore,　traffic　accidents　prediction　is　being　of　great　significance　to　alleviate　the　safety　pressure　of　regional　traffic　management.　Nevertheless,　the　existing　studies　has　yet　reached　a　consensus　on　the　scientific　and　feasible　modeling　method　for　traffic　safety　management,　the　improvement　of　traffic　safety　efficiencies　has　also　rarely　discussed　in　traffic　accidents　prediction.　This　paper　fills　the　gap　by　promoting　a　novel　data-driven　decision　model　for　traffic　accidents　prediction,　which　is　constructed　by　the　extended　belief　rule-based　system　(EBRBS)　with　considering　the　improvement　of　traffic　safety　efficiencies.　Hence,　the　new　traffic　accident　prediction　model　consists　of　two　components:　1)　safety　efficiencies　evaluation　modeling　with　considering　meta-frontier　and　group-frontier　to　evaluate　the　current　traffic　safety　management,　which　are　also　defined　to　improve　safety　efficiencies　evaluation　by　the　adjustment　of　inputs　and　outputs;　2)　extended　belief　rule　base　(EBRB)-based　modeling　for　traffic　accidents　prediction　by　considering　the　improvement　of　traffic　safety　efficiencies,　where　the　effective　efficiencies　of　traffic　management　inputs　and　outputs　are　utilized　to　predict　the　future　number　of　traffic　accidents.　The　effectiveness　of　the　proposed　model　is　verified　by　using　traffic　management　data　from　31　Chinese　provinces　during　2003–2020.　Experimental　results　demonstrate　that　the　model　can　offer　powerful　reference　value　in　the　traffic　accidents　prediction　process,　which　help　to　achieve　the　relatively　effective　efficiencies　of　traffic　safety.　©　2022　Elsevier　Ltd