基于量子行為粒子群優(yōu)化方法的隨機(jī)規(guī)劃算法研究.pdf

1、江南大學(xué)碩士學(xué)位論文基于量子行為粒子群優(yōu)化方法的隨機(jī)規(guī)劃算法研究姓名：李紅梅申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：須文波20080501—_二二—二二二一一一一——壘!!壘壘曼!AbstractMarkowitzadvancedhisfinancialportfoliomodelandalsosetupthebeginningofrationanalysistofinancialinvestmentAfterthat，With

2、therapiddevelopmentofeconomyandthedeeperresearchiScarriedonbythescholars，ontheotherhand，thecircumstanceofinvestmentbecomescomplexandchangeful，anddemandproposedbytheinvestorsbecomeshigherandhigherSOtheportfoliooptimizatio

3、nmodelturnstobemorecomplexAsalongterminvestoritisnotenoughtostudythestaticmodelsBecausetheriskiSinconsistentwithchangingofeconomicenvironment，theinvestorshouldadjustinvestmentstrategyandnotkeepunalterableportfoliountilth

4、eendtimeofinvestmentThatismyresearch：multistagestochasticoptimization，whichmanagesportfolioinconstantlychan百ngfinancialmarketsbyperiodicallyrebalancingtheassetportfoliotoachievereturnmaximizationandriskminimizationTheres

5、earchofportfoliooptimizationiSanuncertainphenomenonItbecomesmoreandmoredifficultinresolvingitwithtraditionalmathematicprogrammingHoweverwiththerecentdevelopmentincomputersandrenovationalgorithm，manycomplexoptimizationpro

6、blemscanbesolvedbyitIntheworld，artificialintelligencemethods，suchasGAandartificialneuralnetworkareintroducedtosolvethefinancialportfolioproblemwhiletraditionalmathematicsmethodsarebecomingmoreandmoredifficultInthispapeLT

7、heMarkovitzportfoliomodeliSfirstlystudiedwhichiSembedintoamultistagestochasticoptimizationmodelThenadecisionmakingprocessisproposedwhichissolvedbyQuantum：behavedParticleSwamOptimization(QPSO)Algorithm，ParticleSwamOptimiz

8、ation(Pso)，andGeneticAlgorithm(GA)tosolvemultistageportfoliooptimizationproblemInordertoevaluatetheperformanceofthesethreealgorithmsonmultistagefinancialoptimization，experimentswerecarriedoutWeeklyclosingpriceofS&P100Ind

9、exanditscomponentstocksfrom1January2000to31December2005werecollectedneperformanceofthemdemonstratedthatthePSOconvergesmostrapidlyinearlystageoftherunning，butmayencounterprematureconvergenceandthereforeonlyfindoutsub—opti

10、maTheGAhastheslowestconvergenceratethantheQPSOandPSObutitencountersprematureconvergenceIessfrequentlythanthePSOTheGAsslowconvergenceratemaycausethepopulationnottoconvergetoapointinthesearchspacewhentherunningisoveECompar

11、ingwiththeothertwoalgorithms，theQPSOcanconvergerapidlyandsearchoutthegiobaloptimamostfrequentlyIftheMarkowitzportfoliomodeliSsetup，anumberofcovariancesneedtobeevaluated，whichneedthehugecomputercapacitytosatisfyitInordert

12、educethecomputationalburden，asingleindexmodleisintroducedThenadecision—makingprocessisproposedwiththethreealgorithmsTheperformanceofthemdemonstratedthatQPSOalsohasagoodsearchingabilityKeywords：StochasticProgramming；Asset