Microarraytechnologycanbeemployedtoquantitativelymeasuretheexpressionofthousandsofgenesinasingleexperiment.Ithasbecomeoneofthemaintoolsforglobalgeneexpressionanalysisinmolecularbiologyresearchinrecentyears.Thelargeamountofexpressiondatageneratedbythistechnologymakesthestudyofcertaincomplexbiologicalproblemspossible,andmachinelearningmethodsareexpectedtoplayacrucialroleintheanalysisprocess.Inthispaper,wepresentourresultsfromintegratingtheself-organizingmap(SOM)andthesupportvectormachine(SVM)fortheanalysisofthevariousfunctionsofzebrafishgenesbasedontheirexpression.Themostdistinctivecharacteristicofourzebrafishgeneexpressionisthatthenumberofsamplesofdifferentclassesisimbalanced.WediscusshowSOMcanbeusedasadata-filteringtooltoimprovetheclassificationperformanceoftheSVMonthisdataset.