时间序列相空间重构数据挖掘方法及其在证券市场的应用

《时间序列相空间重构数据挖掘方法及其在证券市场的应用》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、湖南大学博士学位论文时间序列相空间重构数据挖掘方法及其在证券市场的应用姓名：陈佐申请学位级别：博士专业：管理科学与工程指导教师：谢赤20070209坠塑童型堡耋旦茎丝鍪堡堡堡童兰垒茎奎垩耋皇丝竺窒里常的分类和预测。以决策树的分类标识为指导构建交易策略，在证券市场上进行了应用。结果表明，尽管在股市大势呈现下降趋势的情况下，应用分类标识为指导的交易策略仍能获得较高的收益。本文应用相空间重构技术将时间序列分割成长度相同的子序列集合，并将其映射到多维特征空间，从而将有序的时间序列一维数据挖掘问题转换成为多维空间的无

2、序数据集合的挖掘问题。本文的研究不仅为金融时间序列分析提供了新的方法，也为数据挖掘技术提供了新的研究思路。关键词：相空间重构；时间序列；数据挖掘；小波聚类；EM算法；异常Ⅱ博士学位论文AbstractFinancialmarketsetsupaconnectionbetweenthedemandsofinvestmentandfundmg．Itcouldresolvethecontradictionbetweensupplyanddemandofcapitaleffectively．Analysismeth

3、odsofinvestmentarealwaystheresearchinghotspotoffinancialfield．Withtherapiddevelopmentsofthefinancialmarket,therecomeslotsofcreationandprogressininvestmentanalysis．Traditionaltimesiresmodelshavetwodisadvantages，whichcouldnotbeavoided．Theoneisthatitdependson

4、severalhypothesisconditions．TheotheristhatapplyingoverallfixedmodeltodescribetimeconomicorcommercialtimeseriessWoetures，whicharochangedwithtin'le$gradually,isnotperfectlyapplicable．Withthepopularizationofinformationtechnologyinfinancialfieldandsignificanti

5、mprovementofpeople’Sabilityofcollectingdata,largeamountsofdatawereaccumulated，whichweTefullofabundantinformation,whiletherapiddevelopmentoffinancialmarket．DataminingprovidesUSnewdirectionstoanalyzefinancialtimeseries．Basedonphasespacereconstruction,Thispap

6、ertooktimeseriesaSresearchingobjeettopresenttimeseriesdataminingmethodsandapplythesemethodstofinancialmarket，inordertofmdtheimplicitrules，patternsandknowledge，SOastoprovidenewdirections，methodsandaccessorialinformationtomarketanalysisandinvestmentdecision．

7、Consideringtheresearchingbackground，thispaperdiscussedtheassociatedresearchofdataminingtechnology,timeseriesandfinancialtimeseriesdatamining，separately．Asfollowing，thebasictheoryandmethodsofphasespacereconstructionwereanalyzedindetails．Alloftheseprovidedth

8、etheoreticalbasisandtechnicalfeasibilitytotimeseriesdataminingbasedonphasespacereconstruction．Aftercontrastingthedifferentmeansoftimeseriespattemmining，WepointedouttheproblemoftimeseriesdataminingframeworkTSD