资源描述：
《fuzzy transforms method in prediction data analysis.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、FuzzySetsandSystems180(2011)146163www.elsevier.com/locate/fssFuzzytransformsmethodinpredictiondataanalysisFerdinandoDiMartinoa,VincenzoLoiab,SalvatoreSessaa,∗aUniversitàdegliStudidiNapoliFedericoII,DipartimentodiCostruzionieMetodiMatematiciinArchitettur
2、a,ViaMonteoliveto3,80134Napoli,ItalybUniversitàdegliStudidiSalerno,DipartimentodiMatematicaeInformatica,ViaPonteDonMelillo,84084Fisciano,ItalyAvailableonline25November2010AbstractWepresentapredictionmethodbasedonFuzzytransformsforforecastingproblemsandw
3、ecompareitwiththewellknownWangMendelsone.AnothercomparisonismadeontheLocalLinearWaveletNeuralNetworkmethodviaforecastingtimeseries.WeapplytheseconceptstoaMackeyGlasschaotictimeseriesandtothemonthlyNAOclimaticindextimeseries.©2010ElsevierB.V.Allrightsres
4、erved.Keywords:Fuzzypartition;Fuzzytransform;Fuzzyruleset;Forecastingtimeseries1.IntroductionPredictionandforecastingmodelsarestatisticalmethodsusedtoestimatethefuturetrendofavariablestartingfromnumericaldatasets.Thesemodelsareusedinmanypracticalproblem
5、sincludingrealtimesignalprocesses,currencyforecasting,transportplanning,landuseforecasting,weatherforecasting,etc.[1,4,5,15,1719,21,27,28,42,62].Intheliteraturethereexistsalargequantityofsoftcomputingforecastingmethodsinvolvingfuzzypartitionoftheinputou
6、tputvariablesdomainandfuzzyrulesextraction[9,10,46,50,51,55,58,62],multi-levelneuralnetworkapplications[11,13,14,31,3639,56,59,64]andgeneticalgorithmapplications[25,26,37,40,41].Thesoftcomputingforecastingmethodsarealsousedtodescribeimprecisionanduncert
7、aintyinthedataandthewellknownWangMendelmethod[62]belongstothiscategory.Itgeneratesfuzzyrulesbyanumericaldatasetandusesadefuzzificationprocessforthedeterminationofamappingfromtheinput-variablesspaceintotheoutput-variablesspace.Thismethodcombinesthenumeric
8、aldatawithalinguisticinterpretationoftheproblemgivenbyanexpert.Thismethodhasbeenusedinfuzzyrule-setgeneration,datasetsclassificationandfeatureselectionproblems[2,3,6,8,12,16,32,33,47,48,52,53,57,60,63].Basedonthegranularityofthepr
