神经网络实证方法对我国资本市场有效性及非线性的检验

A--------------------------------------------1------------------------------2efficientmarket--------5----------------------------------------7------------------------10--12-------------------------------12-------------------------------19--------------------------------------------21-=====------------==-------------------------23110Efficiency212343Samulsen,1965Roberts,1967Weak-fromEfficiencySemistrong-fromEfficiencyStrong-fromEfficiencyWest,1975Tobin,1978Arrow-Debreu1120014efficientmarket[]HaimLevy20002EfficientmarketEMT1.weakformoftheEMTrandomwalk32[]HaimLevy2000152.semistrongformofEMTfundamentalanalysis3.strongformofEMT37464[]HaimLevy2000170HaimLevy20003-4519931995199326199319941995561720013-4520017Bachelier1900Roberts1959Osborne1959GrangerMorgenstern1963Fama1965Fama,Lume19668Solink1973Merton1980French1980Keim1983Gultekin1983JaffeEsterfield1984Lehmann199091012ABH34AB111.20005232001723284125.120005232001723(Kurtosis)4.83(Skewness)-0.85JB73.7χ2=5.992.1BP23SIGMOID43512331363.mse=10e-30.00250.320084965.24.5.35.15.214d+1(r2,Ir2,I-1)×(r1,Ir1,I-1)0d=d=d(r2,Ir2,I-1)×(r1,Ir1,I-1)0p=Td/Tdr2r1pT198Td5.3(r2,Ir2,I-1)×(r1,Ir1,I-1)0(r2,Ir2,I-1)×(r1,Ir1,I-1)015015p=Td/T=19815/198=0.924292.42%7p=Td/T=837/83=0.915791.57%A.msemse9.8638e-004B.3713pp=T1[(1r1,ir2,i)/r1,i]16p=1-90.78/198-37161=0.44p=1-37.5854/(83-13)=0.463144%46%6.C1C2C3C0C1C2C3C2C3C1C2C3C1C3C1C25.4175.4123451819971994121220028159089060303445.5195.5959-9/59=84.75%629-6/29=79.31%80%1994121220028154880%90%205.66-16-12109242284310915708475890310793191.57%22[1][]E/1999[2][]HaimLevy20001[3]2001[4]2000[5]2001[6]2000[7]BP2002.165-88[8]“BP20011[9][J].19976127-130[10]“RBF”2000.74483-486[11]Bailey,DavidB.,Thompson,DonnaM.andFeinstein,JeraldL.(1988)“OptionTradingusingneuralnetworks”,inJ.HeraultandN.Giamisasi(eds),ProceedingsoftheInternationalWorkshop,NeuralNetworksandtheirApplications,Neuro-Nimes,15-17November1988,pp.395-402.[12]Baba,N.andKozaki,M.(1992)“Anintelligentforecastingsystemofstockpriceusingneuralnetworks”,inInternationalJointConferenceonNeuralNetworks-IJCNN,Baltimore,Maryland,June1992,Vol.I,pp.371-77.[13]Blayo,F.andDemartines,P.(1991)“Dataanalysis:HowtocompareKohonenneuralnetworkstoothertechniques”,inProceedingsoftheIWANN91Granada,Spain,Springer-Verlag,pp.469-76.23[14]DeGooijer,J.G.(1989)“Testingnonlinearitiesinworldstockmarketprices”,EconomicsLetters,31,31-5.[15]Falma,E.(1970)“Efficientcapitalmarkets:Areviewoftheoryandempiricalwork”,JournalofFinance,25,383.[16]Ferson,W.A.andHarvey,C.R.(1991)“Sourcesofpredictabilityinportfolioreturns”,FinancialAnalystsJournal,May-June,49-56.[17]Foster,B.F.etal.,(1991)“Neuralnetworkforecastingofshortnoisytimeseries”,presentedattheORSATIMSNationalMeeting,May.[18]Hoptroff,R.G.(1993)“Theprinciplesandpracticeoftimesseriesforecastingandbusinessmodelingusingneuralnetworks”,NeuralComputingandApplications,1.[19]Refenes,A.N.,Zapranis,A.D.andBentz,Y.(1993)“Modelingstockreturnswithneuralnetworks”,inA.N.Refenes(ed.),NeuralNetworksintheCapitalMarkets,Proc.NnCM”93,LondonBusinessSchool,November.[20]Schoenenburg,E.(1990)“Stockpricepredictionusingneuralnetworks:Aprojectreport”,Neurocomputing,2.[21]WilliamA.BarnettApostolosSerletisMartingalsnonlinearityandchaosJournalofEconomicDynamics&Control24(2000).24