Predicting Breast Cancer Survivability Using Data Mining Techniques

《Predicting Breast Cancer Survivability Using Data Mining Techniques》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、PredictingBreastCancerSurvivabilityUsingDataMiningTechniquesAbdelghaniBellaachia,ErhanGuvenDepartmentofComputerScienceTheGeorgeWashingtonUniversityWashingtonDC20052{bell,eguven}@gwu.eduAbstractrelationshipoftheassociation.DatadrivenstatisticalresearchisbecomingacommoncomplementtomanyIn

2、thispaperwepresentananalysisofthepredictionscientificareaslikemedicineandbiotechnology.Thisofsurvivabilityrateofbreastcancerpatientsusingtrendisbecomingmoreandmorevisibleasinthedataminingtechniques.ThedatausedistheSEERstudiesofHoustonetal.[5]andCiosetal.[6].Public-UseData.Thepreprocess

3、eddatasetconsistsInthispaper,wepresentdataminingtechniquesof151,886records,whichhavealltheavailable16topredictthesurvivabilityrateofbreastcancerfieldsfromtheSEERdatabase.Wehaveinvestigatedpatients.Inourstudy,wehaveusedtheSEERdatathreedataminingtechniques:theNaïveBayes,theandhaveintrodu

4、cedapre-classificationapproachback-propagatedneuralnetwork,andtheC4.5thattakeintoaccountthreevariables:SurvivalTimedecisiontreealgorithms.SeveralexperimentswereRecode(STR),VitalStatusRecode(VSR),andconductedusingthesealgorithms.TheachievedCauseofDeath(COD).predictionperformancesarecomp

5、arabletoexistingThispaperisorganizedasfollows.Thenexttechniques.However,wefoundoutthatC4.5sectionreviewsrelatedwork.Section3givesthealgorithmhasamuchbetterperformancethanthemethodologyusedtoconductthepredictionanalysis.othertwotechniques.ExperimentalresultsarepresentedinSection4.Conclu

6、sionandfutureworkaregiveninthelastKeywords:Breastcancersurvivability,datamining,section.SEER,Weka.2.RelatedWork1.IntroductionAliteraturesurveyshowedthattherehavebeenToday,intheUnitedStates,approximatelyoneinseveralstudiesonthesurvivabilitypredictioneightwomenovertheirlifetimehasariskof

7、problemusingstatisticalapproachesandartificialdevelopingbreastcancer.Ananalysisofthemostneuralnetworks.However,wecouldonlyfindafewrecentdatahasshownthatthesurvivalrateis88%studiesrelatedtomedicaldiagnosisandsurvivabilityafter5yearsofdiagnosisand80%after10yearsofusingdataminingapproac