It　is　necessary　to　obtain　corresponding　solutions　to　evaluating　the　fitness　of　all　individuals　of　every　generation　of　the　population　and　to　analyze　the　solutions　by　using　Genetic　Algorithm.　When　the　scale　of　problem　is　large,　the　calculation　of　genetic　algorithm　will　be　so　enormous　that　it　ean　not　be　used　in　practice.　However,　a　new　method　called　empirical　genetic　algorithm　is　proposed　in　the　paper.　It　decrease　the　number　to　analyze　the　solution　and　increase　the　efficiency　of　the　genetic　algorithm,　in　which　the　fitness　of　most　individuals　of　every　generation　of　the　population　are　estimated　by　the　empirical　Neural　Network.　The　calculation　results　from　six　classical　test　functions　show　that　the　method　is　efficient.