资源描述：
《Dynamic Ridge Polynomial Neural Network Forecasting the univariate》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、ExpertSystemswithApplications38(2011)3765–3776ContentslistsavailableatScienceDirectExpertSystemswithApplicationsjournalhomepage:www.elsevier.com/locate/eswaDynamicRidgePolynomialNeuralNetwork:Forecastingtheunivariatenon-stationaryandstationarytradingsignalsa,⇑bcRozaidaGhazali,Abir
2、JaafarHussain,PanosLiatsisaInformationTechnologyandMultimediaFaculty,UniversitiTunHusseinOnnMalaysia,MalaysiabSchoolofComputingandMathematicalSciences,LiverpoolJohnMooresUniversity,UKcSchoolofEngineeringandMathematicalSciences,CityUniversity,London,UKarticleinfoabstractKeywords:Th
3、ispaperconsidersthepredictionofnoisytimeseriesdata,specifically,thepredictionoffinancialsig-DynamicRidgePolynomialNeuralNetworknals.AnovelDynamicRidgePolynomialNeuralNetwork(DRPNN)forfinancialtimeseriespredictionisFinancialsignalspresentedwhichcombinesthepropertiesofbothhigherorderan
4、drecurrentneuralnetwork.InanHigherorderneuralnetworkattempttoovercomethestabilityandconvergenceproblemsintheproposedDRPNN,thestabilitycon-TimeseriespredictionvergenceofDRPNNisderivedtoensurethatthenetworkpossesauniqueequilibriumstate.Inordertoprovideamoreaccuratecomparativeevaluat
5、ionintermsofprofitearning,empiricaltestingusedinthisworkencompassnotonlyonthemoretraditionalcriteriaofNMSE,whichconcernedathowgoodtheforecastsfittheirtarget,butalsoonfinancialmetricswheretheobjectiveistousethenetworkspredic-tionstogenerateprofit.Extensivesimulationsforthepredictionofo
6、neandfivestepsaheadofstationaryandnon-stationarytimeserieswereperformed.TheresultingforecastmadebyDRPNNshowssubstantialprofitsonfinancialhistoricalsignalswhencomparedtovariousneuralnetworks;thePi-SigmaNeuralNetwork,theFunctionalLinkNeuralNetwork,thefeedforwardRidgePolynomialNeuralNet
7、work,andtheMultilayerPerceptron.SimulationresultsindicatethatDRPNNinmostcasesdemonstratedadvanta-gesincapturingchaoticmovementinthefinancialsignalswithanimprovementintheprofitreturnandrapidconvergenceoverothernetworkmodels.Ó2010ElsevierLtd.Allrightsreserved.1.Introductiontionunitson
8、ly.Theutilizationofhigherorderter
