In　the　present　paper,　we　present　a　new　way,　named　hybrid　method,　for　selecting　specific　genes　from　a　gastric　cancer　dataset　of　29　samples　from　Japan.　Finally,　10　related　genes　have　been　picked　out　and　2　of　them　are　found　to　be　the　same　with　the　reported　by　Kim＇.　The　hybrid　method　combines　the　advantages　of　both　filter　and　wrapper　method　in　feature　selection.　In　the　last,　we　employ　the　support　vector　machine　(SVM),　artificial　neural　network　(ANN)　and　K-nearest　neighbors　(KNN)　to　evaluate　the　validity　of　our　selected　genes,　which　indicates　that　it　is　effective.