This　paper　applies　the　XGBoost　method　to　construct　a　predictive　model　for　the　risk　of　type　2　diabetes　which　based　on　the　physical　examination　data.　The　paper　takes　the　real　physical　examination　records　of　the　same　batch　of　people　in　a　health　check-up　center　from　2010　to　2015　as　the　data　source,　and　evaluates　the　feature　importance.　Finally,　28　characteristic　variables　are　selected　as　the　model　input,　and　a　phase　is　obtained.　Compared　with　other　common　classification　algorithms,　the　prediction　model　with　higher　prediction　accuracy　and　stronger　generalization　ability　has　certain　clinical　reference　value　for　the　risk　prediction　of　type　2　diabetes.