基于LIBSVM的葡萄酒品质评判模型(1)

235201010()JournalofSichuanUniversityofScience&Engineering(NaturalScienceEdition)Vol123No15Oct12010:2010207207:(2008RCYJ09);(07ZR41):(19812),,,,,:167321549(2010)0520530203LIBSVM,(,643000):,178,13,,LIBSVM,98%,:;;;:TP183:A,,,,,,,,,,,11.1UCIwine[1],,:178,13,,13,Alcohol(g/L),Malicacid(g/L),Ash(g/L),Alcalinityofash(g/L),Magnesium(g/L),Totalphenols(g/L),Flavanoids(g/L),Nonflavanoidphenols(g/L),Proanthocyanins(g/L),Colorintensity(g/L),Hue(g/L),OD280/OD315ofdilutedwines(g/L),Proline(g/L):159,271,348178,13,31.2[2](supportvectormachines,SVM),VC,(,),,(),,178,,,VSM[3]22.1SVM:(1),,,;(2),,(),,:(1),n,,(2)(3),,(4),,,,,iyi[(w.xi)+b]-1+iE0i=1,2,n(1),w2Fck(2)0,(1)(2)F()=ni=1i,=1,(1)(w,)=12(ww)+C(ni=1i)(3),w2,ni=1i,C,,C,,C,,C[4],K,K,K(K-1)/2SVM,minwij,bij,ij12(wij)Twij+Cni=1iji(4)(wij)T(xi)+bijE1-ijiyi=i(5)(wij)T(xi)+bijE-1+yii(6)ijiE0j=1,2,n(7)12(wij)Twij2wij,,Cni=1iji,sgn(wij)T((xi)+bij)xi,i,,xj,j,x,x2.2(1),LIBSVM[5],,traindatatestdata,,2/3,1/3wine.scalewinetrainwinetest(2)SVM,,sigmoid,,[6]RBFK(x,y)=e-gx2-y2,,(3)SVM,SVMCg,Cg[7],wine2train.txt.model(4)135235:LIBSVMwinetrainwinetrain.tx.modelwinetrain.txt.outwine.txt.out60,20,20,20,1,1/602.31,Cg,,a,,f,,g,C,,Cg,9818%,13SVM,,,,,,,[8]:[1]FrankA,AsuncionA.UCIMachineLearningRepository[DB/OL].UniversityofCalifornia,Irvine,SchoolofInfor2mationandComputerSciences.[2].[M].:,2006.[3]CortezP,CerdeiraA,AlmeidaF,etal.Modelingwineprefer2encesbydataminingfromphysicochemicalproperties[J].DecisionSupportSystems,2009,47(4):5472553.[4]FrankA,AsuncionA.UCIMachineLearningRepository[DB/OL].UniversityofCalifornia,Irvine,SchoolofInformationandComputerSciences.[5].[R/OL].@%BBP%A4%F1%B8%FB.pdf.[6],.[M].:,2006.[7]HsuChih2Wei,ChangChih2Chung,LinChih2Jen,APracticalGuidetoSupportVectorClassification[K/OL].[8],,,.[J].,2010,34(2):1332136.ModelofWineQualityIdentificationBasedonLIBSVMGAOYuan2yuan,LIUQiang2guo(SchoolofScience,SichuanUniversityofScience&Engineering,Zigong643000,China)Abstract:Numerousandcomplexingredientofwine,isanimportantbasisforthequalityofwine.Afterprocessingandanalyzingchemicalanalysisdataof178winesamples,whichcontains13properties,themodelofwinequalityidentificationbasedonsupportvectormachineisproposedinthisarticle.BymeansofLIBSVM,thecomplexhighdimensionwinepropertydataisanalyzed,processed,optimizedandinterpreted.Theprecisionofresultaboutclassifyingthewineis98%,therebyitoffersatheoreticalbasisforidentifyingthewinequalityrapidlyandefficiently.Keywords:supportvectormachine;dernelfunction;radialbasisfunction;penaltycost235()201010