資源描述:
《基于灰色系統(tǒng)理論的數(shù)據(jù)挖掘及其模型研究》由會員上傳分享,免費在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、北京交通大學(xué)碩士學(xué)位論文基于灰色系統(tǒng)理論的數(shù)據(jù)挖掘及其模型研究姓名:徐進(jìn)華申請學(xué)位級別:碩士專業(yè):信息管理指導(dǎo)教師:姚家奕20090601中文摘要摘要:數(shù)據(jù)挖掘作為一個從大規(guī)模海量數(shù)據(jù)中提取隱含的有價值信息和知識的過程,已經(jīng)被人們廣泛地應(yīng)用于社會、經(jīng)濟(jì)、生產(chǎn)、生活的各個領(lǐng)域。但是數(shù)據(jù)挖掘有其局限性:數(shù)據(jù)量必須要大到足夠辨認(rèn)出期望的關(guān)系?;疑到y(tǒng)理論作為一種研究少數(shù)據(jù)、貧信息不確定性問題的新方法,恰恰可以彌補數(shù)據(jù)挖掘的這一缺陷。雖然灰色系統(tǒng)理論本身也存在很多缺陷,但是數(shù)據(jù)挖掘領(lǐng)域的很多技術(shù)卻可以克服這
2、些問題。因此,可以將灰色系統(tǒng)理論和數(shù)據(jù)挖掘技術(shù)結(jié)合起來,兩者優(yōu)勢互補,建立基于灰色系統(tǒng)理論的數(shù)據(jù)挖掘體系和灰色數(shù)據(jù)挖掘模型?;谶@樣的思想,研究人員提出了幾種灰色神經(jīng)網(wǎng)絡(luò)組合模型,但這些模型都只是灰色系統(tǒng)理論和人工神經(jīng)網(wǎng)絡(luò)的簡單組合,而且都只能對一個數(shù)據(jù)序列做預(yù)測,不能考慮同時對多個相關(guān)序列進(jìn)行預(yù)測的情況。本文以此為出發(fā)點,將多個序列之間的相互關(guān)系及GM(1,1)模型預(yù)測值和實際值之間的偏差關(guān)系綜合到BP神經(jīng)網(wǎng)絡(luò)模型中來考慮,建立了灰色神經(jīng)網(wǎng)絡(luò)多序列預(yù)測模型;并將已有的灰色神經(jīng)網(wǎng)絡(luò)組合模型和灰色神經(jīng)
3、網(wǎng)絡(luò)多序列預(yù)測模型應(yīng)用于股票價格預(yù)測,驗證了灰色神經(jīng)網(wǎng)絡(luò)多序列預(yù)測模型的可行性和優(yōu)越性。關(guān)鍵詞:數(shù)據(jù)挖掘;灰色系統(tǒng)理論;灰色數(shù)據(jù)挖據(jù);BP神經(jīng)網(wǎng)絡(luò)分類號:TP311ABSTRACTABSTRACT:Asacourseofextractingvaluableimpliedinformationandknowledgefromlmgeamountofdata,DataMininghasbeenwidelyusedinsocie壩economy,production,andallarea8ofourlif
4、e.ButtherearelimitationsinDataMiningthattheamountofdatamustbelargeenoughtoidentifytheantiopantrelationships.Asanewmethodtoresearchproblemsthathavelessdata,poorinformationanduncertainty,GreySystemTheorycanmakeupfortheshortcomingsofDataMinin參AlthoughGrey
5、SystemTheoryitselfhasmanydefccts,manytechnologiesintheareaofDataMiningCanovercometheseproblems.’There如m,wecmcombineGreySystemTheoryandDataMiningTechnologiestogether.TheycomplementeachotherandthenwecanbuildsystemandmodelsbasedonDataMiningandGreySystemTh
6、eory.Basedonthisthinking,researchershaveproposedseveralcombinedmodelsofgreyneuralnetwork.ButthesemodelsareonlythesimplecombinationsofGreySystemTheoryandArtificialN刪NetworkandCanonlydoforecastonaseriesofdata,meycan’tbeusedtoforecastonanumberofrelatedseq
7、uences.SothispaperintegratesthemutualrelationshipsbetweenseveralsequencesandthedeviationrelationshipbetweenthepredictionvalueandtheactualvalueforecastedbytheGM(I,1)modelintotheBPNepalNetworkModelandbuildstheGreyNeuralNetworkMulti—seriesPredictionModel.
8、AndthispaperforecaststhestockpriceusingthecurrentcombinedmodelofGreyNeuralNetworkandtheGreyNe町a(chǎn)lNetworkMulti—seriesPredictionModelthatwehavebuiltandverifiesthefeasibilityandsuperiorityoftheGreyNepalNetworkMulti—seriesPredictionModel.KEY