Modern　medicine　has　made　immense　advance　in　the　medical　diagnostic　techniques,　which　benefits　the　professors　an　increasingly　growing　number　of　physical　exam　projects　with　higher　accuracy.　The　result　is　that,　nowadays,　it　has　the　promise　to　make　precise　disease　prevention　through　monitoring　the　indexes　of　the　patient＇s　biochemical　indicators.　In　the　past,　determining　the　biochemical　indicators　for　a　disease　was　a　crucial　issue　and　was　mainly　carried　out　with　the　domain　experts＇　experience　and　professional　knowledge.　However,　this　may　be　time-consuming　and　may　have　added　some　subjectivity　and　partial　opinions.　As　a　result,　there　has　been　growing　interest　in　targeting　the　biochemical　indicators　fast　and　right　with　the　aid　of　computer　science.　In　this　paper,　we　adopt　the　feature　selection　algorithm　for　Clusters　(FSC)　to　mine　a　set　of　the　chosen　SGA　indicators　aiming　to　find　the　ones　with　the　most　characteristic　strength.　For　the　stage　of　feature　selection,　we　cluster　the　features　and　find　the　top　variable　which　shares　the　most　discriminative　information.　Finally,　to　exam　the　efficacy　of　the　FSC,　four　widely　used　classifiers　are　implemented　in　the　experiment.　The　results　show　that　for　the　prediction　of　SGA,　the　FSC　can　reach　a　prediction　precision　of　78.81%,　which　validates　the　feasibility　of　applying　this　algorithm　to　mining　the　indicators　for　SGA.