资源描述：
《数据流上的预测聚集查询处理算法》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、1000-9825/2005/16(07)1252©2005JournalofSoftware软件学报Vol.16,No.7∗数据流上的预测聚集查询处理算法+李建中,郭龙江,张冬冬,王伟平(哈尔滨工业大学计算机科学与技术学院,黑龙江哈尔滨150001)ProcessingAlgorithmsforPredictiveAggregateQueriesoverDataStreams+LIJian-Zhong,GUOLong-Jiang,ZHANGDong-Dong,WANGWei-Ping(InstituteofComputerScienc
2、eandTechnology,HarbinInstituteofTechnology,Harbin150001,China)+Correspondingauthor:Phn:+86-451-86415827,E-mail:lijzh@hit.edu.cn,http://db.cs.hit.edu.cnReceived2004-05-17;Accepted2005-02-03LiJZ,GuoLJ,ZhangDD,WangWP.Processingalgorithmsforpredictiveaggregatequeriesoverdata
3、streams.JournalofSoftware,2005,16(7):1252−1261.DOI:10.1360/jos161252Abstract:Itisveryimportantinalotofapplicationstoforecastfuturetrendofdatastreams.Forexample,usingpredictivequeriestoasensornetworkformonitoringenvironment,observerscanforecastfutureaveragetemperatureandh
4、umidityintheareacoveredbythenetworktodetermineabnormalevents.Recentworksonqueryprocessingoverdatastreamsmainlyfocusedonapproximatequeriesovernewlyarrivingdata.Tothebestoftheknowledge,thereisnothingtodateintheliteratureonpredictivequeryprocessingoverdatastreams.Adoptingmu
5、ltivariablelinearregression,apredictivemathematicalmodelforforecastingtheaggregatevalueoverdatastreamsisfirstproposed.Then,basedonthemodel,apredictiveaggregatequeryprocessingmethodoverdatastreamsisproposedinthepaper.Whenthefrequencyofforecastfailingisgreaterthanapredefin
6、edthreshold,anadaptivestrategyforthepredictivemathematicalmodelisproposed.Amathematicalmodelthatcharacterizestheaffectsoftheupdatingcycleofslidingwindowanddatastreamrateonpredictiveaccuracyisalsopresented.Analyticalandexperimentalresultsshowthattheproposedmethodisveryeff
7、ective,andtheproposedalgorithmshavehigherperformanceandprovidebetterpredictionofaggregatevaluesoverdatastreamstousers.InexperimentstheTPC-HdataandoceanairtemperaturedatameasuredbyTAO(tropicalatmosphereocean)areusedtoconstructdatastreams.Keywords:datastream;futuredatawind
8、ow;multivariablelinearregression;predictiveaggregatequeries摘要:实时数据流未来趋势的预测具有重要的实际应用意义.例如,在环境监测传感器网络中,通过
