時(shí)間序列異常模式挖掘關(guān)鍵技術(shù)研究.pdf

1、西安科技大學(xué)碩士學(xué)位論文時(shí)間序列異常模式挖掘關(guān)鍵技術(shù)研究姓名：周鑫申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：李愛(ài)國(guó)201106Subject:StudyofKeyTechnologiesonAbnmalPatternMininginTimeSeriesSpecialty:ComputerApplicationTechnologyName:ZhouXin(Signature)Instruct:LiAiguo(Signature)AB

2、STRACTStudyofkeytechnologiesonabnmalpatternminingintimeserieshasimptantacademicvaluepracticalsignificance.Indertodealwiththecontinuitynonlinearhighdimensionoftimeseriesthisthesisaimstoexpleanewwaytomineabnmalpatternoftim

3、esseries.Themethodsofabnmaldetectionbasedonclassficationdimensionalityreductionofactersattributesarestudiedwhichareappliedtoanomalydetectionofcoalminegasdatas.Atthesametimethesolutionsdefinitionoflayoutofmultisenssystema

4、reproposed.Anabnmaldetectionsoftwaresystemofgasbasedontimeseriesdataminingisdeveloped.Themaincontentsareasfollows:Aninverseclassificationmethodofquantitativeattributeswaspresentedwhichovercomesthedisadvantageofmostinvers

5、eclassificationalgithmsaddressdiscreteattributes.Thealgithmputsemphasisonanalysisofthetrainingsamplesthetestsamplesthemainideais:firstlyagroupoffeatureattributesareedbyusingfeatureionalgithmthenthequantitativeattributesa

6、rediscretizedbyusingdiscretizationalgithmstheinvertedstatisticsareconstructedonthetrainingsamplesfinallythetestsamplesareanalyzedbyusingtheinvertedstatisticsthemethodisappliedtoabnmaldetectionestimatingthemissingvalues.E

7、xperimentalresultsonIRISEcolidatasetsshowthattheaccuracyofclassificationtheaveragerelativedeviationthemaximumrelativedeviationarebetterthanKNN(kNearestNeighbs)ISGNN(IterationSelfGeneratedNeuralwk).Indertoovercomethelowac

8、curacyofanomalydetectionbasedonmanifoldlearningmethods.TwosupervisedmanifoldlearningalgithmsareproposednamedPSLLE(SupervisedLocallyLinearEmbeddingbasedonProblitity)PQSLLE(QuickSupervisedLocallyLinearEmbeddingbasedonProbl