独立成份分析方法在股票分析中的应用

14730(ICA)(BSS)(ICs)(PCA)[1](SVD),PCAICAICAICA[2][3][4][5][6]1ICAICAJuttenHeraultICAICAComonICAICA1.1ICAICAx=As+(1)xA×s(1)A(ICs)sx=As(2)[e-mail](1979)(361005)K;;;;TP319.4A1000-7024(2007)06-1473-04ApplicationofindependentcomponentanalysistoanalysisofstockmarketCHENYu-shan,XIBin(InstituteofPatternRecognitionandIntelligenceSystem,SchoolofInformationScienceandTechnology,XiamenUniversity,Xiamen361005,China)AbstractInordertoovercometheshortcomingofthetraditionalmethods,independentcomponentanalysis(ICA)isusedtoanalysethefactorthathasinfluenceonthetrendofstocksandthereturnsofstocks.Byanalyzingthetrendoftheday-K-linesofthehistoricaldatum,thismethodshowsthefactorthatinfluencesthestockmarketandthereturns.Tosomeextent,itisimportantforconstructingaharmonioussociety.Atthesametime,itisea1474[7](ML)[7](MMI)[7](KICA)[8]ICAICAx[9]01x=wxwwA1wz=Aw=wx=wAs=zs==1zwwxzwx=zswxw=log(3)=(4)(4)=0(3)(4)2(5)v=exp2/2(6)=logcosh12(7)1.2.2ICAwxmaxwx2(8).w=1(9)()ICs[10][11]ICs=1,2,,W=1,2,wxwxwxw=1,2,,G=Wx1.3ICA31=sin/2(10)2=2cos2/4(11)3=.2713(12)31A2ICA3ICAsAsA(1)sA2ICA1010203040506070809010010-110-120-2220-210-120-201020304050607080901003100-1020-220-201020304050607080901001475ICA2.120051130400K=InIn1,10,=1(13)0,22=01exp22=1+0.52(14)Laplace,=0.5exp(15)ICA2.25020041250501010×40010ICAICA10×1010×1010×40045ICA5123456783ICAICA:[1],,.[M].:4642105020100105201002010020100151051050105005010015020025030035040050-520-250-550-550-550-550-550-505010015020025030035040051476,2005.[2],.[M].:,2006.[3],,.[J].,2005,10(2):241-244.[4]SvensenM,KruggleF,BenaliH.ICAoffMRIgroupstudydata[J].NeuroImage,2000,11:24-35.[5],,.ICA[J].,2004,30(9):159-161.[6]ErkkiOjia.Independentcomponentanalysisforfinancialtimeseries[C].AdaptiveSystemsforSignalProcessing,Communica-tions,andControlSymposium,AS-SPCC,IEEE,2000.111-116.[7],,.[J].,2002,28(5):762-772.[8]BachFR,JordanMI.Kernelindependentcomponentanalysis[J].JournalofMachineLearningResearch,2002,(3):1-48.[9]HyvarinenA,OjaE.Independentcomponentanalysis:Algori-thmsandapplications[J].NeuralNetworks,2000,13:411-430.[10]AapoHyvarinen,JuhaKarhunen,ErkkiOja.Independentcompo-nentanalysis[M].NewYork:JohnWileyandSonsInc,2001.[11],,.[J].,2005,26(5):1155-1157.(1465)4:[1],,.[J].,2005,26(11):3021-3023.[2],.[J].,2005,26(12):3259-3260.[3].[J].,2005,31(9):201-203.[4],.[J].,2004,30(6):63-64.[5],.[J].,2005,31(8):63-65.[6].[J].,2005,26(10):2671-2673.[7],.[J].,2005,23(4):72-74.[8],.Domino[J].,2005,31(24):84-86.[9].[J].,2005,26(12):37-38.3PetriNetDSAMRAPWIDWIMRI6CTSAH:[1]YANGJuan,TUDa-wei,CHENGSheng.Optimumdesignofthe360ºrotationobjectsurfaceprofilmeter[J].OpticalTechnique,2003,29(4):402-404.[2],.[J].,2002,13(9):962-965.[3],,,.[J].,2000,40(5):5-8.[4],,.[J].,2000,19(5):804-807.[5],,,.[J].,2003,23(3):272-277.[6],,.360°[J].,2003,29(4):402-404.[7],,,.[J].(),2002,42(4):447-480.[8],,.[J].,2000,20(6):715-719.(1454)