资源描述:
《基于数据挖掘的金融时间序列预测研究与应用.pdf》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、AbstractAbstractWontheNobelPrizeineconomicsin1997,RobertMertonbelievethatthetimevalueofmoney,assetpricingandriskmanagementconstitutesthecornerstoneofmodernfinancialtheory,howintheuncertainenvironmentforinter-temporaloptimalallocationofresourcesisthemostimportantproblem.Withthedevelopmentofco
2、mputersoftware,avarietyofsoftwareandapplicationsarewidelyused,andcontinuedintothebusiness,socialproductionandlife,etc.Financeisanextremelyimportantpartofsocialeconomy.Onfinancialdatafordatamining,itisinalargenumberoftransient,nonlinearandhighsignal-to-noiseratioofuncertainfinancialdatamining
3、valuableinformation.Dataminingtechnologyinthefinancialsectorgraduallyrise.Thetraditionaldataminingtechniquesinthetreatmentofthecommondataperformedwell,buttheprocessingofinstabilityoffinancialtimeseriestoshowsomelimitations.Thusimprovingexistingdataminingtechnologyintheapplicationresearchoffi
4、nancialtimeseriesisparticularlyimportant.Inordertosolvethisproblem,thispaperbasedontheclusteringindataminingresearchasthebreakthroughpoint,themainworkisasfollows:First,inviewofDBSCANclusteringalgorithmcannotdealwiththedatasetsofvarieddensities,combinedwiththeinitialpointofoptimizationandpara
5、meteradaptivemethodforimprovingDBSCANalgorithm.Thispaperproposesanewdatasetcancopewithchangedensityspatialclusteringalgorithmbasedondensity(OS-DBSCAN).Theexperimentalresultsshowthatthenewimproveddensitybasedspatialclusteringalgorithmcandealwithvarieddensitydatasetsforclustering,andaftergivin
6、gtheinitialparametersaccordingtothecharacteristicsandattributesofdatasetsitsownparameteradaptive,andcomparedwiththetraditionalDBSCANalgorithm,thedensityofinitialpointoptimizationandparameteradaptivespatialclusteringalgorithmcanimprovethequalityofclustering.Secondly,inviewofthefinancialtimese
7、riespredictionbasedondataminingproblems,combinedwiththeproposedOS-DBSCANclusteringalgorithmandSVRregressionpredictionalgorithmbasedonparticleswarmoptimization,thispaperproposesahybridalgorithmtocopewithunsteady,nonlinearandhighsignal-to-noiseratioo