灰色神經(jīng)網(wǎng)絡預測模型的優(yōu)化研究.pdf

1、武漢理工大學碩士學位論文灰色神經(jīng)網(wǎng)絡預測模型的優(yōu)化研究姓名：李小燕申請學位級別：碩士專業(yè)：計算機軟件與理論指導教師：袁景凌20090501AbstractGrayneuralnetworkmodelhasbeenarelativelywiderangeofapplicationswiththeindustryneedsforsmallsampleofdatainformationprocessingAlthoughtheneuralne

2、tworkandgraysystemtheoryareappliedwidelyininformationprocessing，theresultsofpredictiondataareunsatisfactoryOfcotlrseestablishedgrayneuralnetworkmodelsoftheintegrationofthesetwomethodshavegooddataprocessingandprediction，a

3、nditcartmakeupthelackofonlyusinganyoneofthemGrayneuralnetworkmodelissimple，iftooptimizethelearningalgorithmandthemodelstructureandperformanceweareabletoachievebeaerresults。Atpresentmanyscholarsdoconsiderableresearchesabo

4、utoptimizing艫ayneuralnetworkandmadeoutsomeachievements，OptimizedgrayneuralnetworkmodelCandealwithuncertaininformationandtechnicaldatatoimproveforecastaccuracyithasbecomeaveryimportanttopic。Atfirst，thepaperstatesthecharac

5、tersofsmallsampledataandthenanalyzesthecomplexityandtheparticularityofestimatingsmallsamples，twomethodsareproposedbycomparisonandvalidate，andthethinkingwhicharebasedon琴夠predictionmodelandneuralnetworktechnologyisadoptedI

6、nthepaperaSGRBFstaticmodelisestablishedonthebasisofRBF(RadialBasisFunction)andGreyModel(O，N)ThemodelCandealwiththepredictionproblemverywell，becauseitmakesuseoftheRBFsgoodabilityofinapproachingnonlinearfunction，andtheaccu

7、racyofGreyModel(O，N)inmakingapredictionofsmallsampledataADGRBFdynamicmodelisalsoestablishedinthepapeLwhichcanselectthebestinitializationconditionsanddynamicidenti掰ngparametersandisfitfordynamicandlong—termdataprediction。

8、WhencombininggreysystemwithRBFneuralnetworklocaloptimizationandconvergenceproblemsarestillexisted，SOgeneticalgorithmisintroducedtoassistthemodelingofgreyneuralnetworkinthispaperFirstlygeneticalgorithmisemployedtosolvethe