基于神經(jīng)網(wǎng)絡和模糊理論的股票指數(shù)收益率預測分析.pdf

1、湖南大學碩士學位論文基于神經(jīng)網(wǎng)絡和模糊理論的股票指數(shù)收益率預測分析姓名：高振坤申請學位級別：碩士專業(yè)：管理科學與工程指導教師：熊正德20050328AbstractSincethestockmarkethasestablishedmorethanonehundredyearsago，howtopredictthestockpriceaccuratelyhasbeenoneofthehottesttopicsthatattractedma

2、nyscholars’andinvestors’atcentionWhiletherealesomanyinfluencingfactorsthataffectthemovementofstockpricesAllthefactorsvaryintheinfluencedegree，timerangeandaffectivemodewhichresultsintheexcepfionalcomplexityofstockpriceSoi

3、tisverydifficulttomakesomeexactforecastofthestockpriceStillmanyscholarandinvestorsinsistonthestudyandpracticetofindthearbiUagechangetoachievetheexcessproceedswhichcomesfromtheaccurateforecastingofstockpricesAndthereareso

4、meforecastingmethodsdevelopedfromdifferentpositiondiversetheorydistinctinveststrategyandpracticeexperienceThispapermainlyputsup、ⅣitltheapplicationofartificialintelligenceandfuzzytheoryintheStockmarketFirstlywepresentamet

5、hodforstockpriceindexforecastingusingneuralnetworkwhichinvolvestheBPmodelandBPGARCHmodelandgivesafurtherdiscussiononthestockpriceforecastingapplicationAndtheShenzhenStockPriceIndexhasbeenappliedforecastingTheempiricalres

6、ultshowsthatcomparingtoBPmodel，theBP—GARCHmodelhasmorefavorablecharacteristicsuchastheconvergencerate，learningabilityforecastingprecisionandestimatingerrorSecondlyAnapproachtoforecastStockIndexbyusinginterpretablefuzzymo

7、delsispresentedinthispaperFuzzyclusteringalgorithmisusedtoidentifiedinitialfuzzymodelThenumberoffuzzyrulesisdeterminedbyclustervaliditymeasure。Similarfuzzysetsmergingmethodisthenappliedtemoveredundancyofthefuzzymodeltoim

8、proveitsinterpretabilityInordertoobtainhighaccuracyyetpreservinginterpretabilityaconstrainedLevenbergMarquardtmethodisutilizedtooptimizefuzzymodelTheproposedapproachisappliedtoShenZhenStockIndexforecastingprocess，andresu