Integrating Frequent Pattern Mining from Multiple Data Domains for Classification

《Integrating Frequent Pattern Mining from Multiple Data Domains for Classification》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、2012IEEE28thInternationalConferenceonDataEngineeringIntegratingFrequentPatternMiningfromMultipleDataDomainsforClassiſcationDhavalPatelWynneHsuMongLiLeeNationalUniversityofSingaporeSingapore{dhaval,whsu,leeml}@comp.nus.edu.sgAbstract—Manyfrequentpatternminingalg

2、orithmshavebeentomineusefulfrequentpatternsforclassiſcation[10],[24],developedforcategorical,numerical,timeseries,orinterval[14],[5],[6].data.However,littleattentionhasbeengiventointegratetheseAstraightforwardmethodtodiscoverheterogenouspatternsalgorithmssoasto

3、minefrequentpatternsinvolvingmultipleistoapplydifferentfrequentpatternminingalgorithmsforthedatadomainsforclassiſcation.Inthispaper,weintroducethenotionofaheterogenouspatternthatcapturestheassociationsdifferentdatadomains,followedbyanexhaustivecombina-amongdiff

4、erentdatadomains.Weproposeauniſedframeworktionsofthediscoveredpatterns.Aquickcalculationrevealsforminingmultipledomainsanddesignaniterativealgorithmthatthisapproachiscomputationallyinfeasible.AsmallcalledHTMiner.HTMinerdiscoversessentialheterogenouspat-datasetw

5、ith10categoricalattributes,20numericalattributes,ternsforclassiſcationandperformsinstanceelimination.This10events,and10daysof10timeseriesdatawouldresultinstanceeliminationstepreducestheproblemsizeprogressivelyinthegenerationof210frequentitemsets[22],220frequent

6、byremovingtraininginstanceswhicharecorrectlycoveredbythediscoveredessentialheterogenouspattern.Experimentsonintervals[9],1010frequenttemporalpatterns[14],and1010tworealworlddatasetsshowthattheHTMinerisefſcientandtimemotifspatterns[15].Thecombinationofthesepatte

7、rnsiscansigniſcantlyimprovetheclassiſcationaccuracy.58oftheorder2,andonlyasubsetofthesepatternsareusefulforclassiſcation.Clearly,weneedamoreintegratedapproachI.INTRODUCTIONtodiscoverusefulpatternsfromdifferentdatadomainsforeffectiveclassiſcation.Manydatabaseapp

8、licationsinvolverecordswithattributesEarlyworksonheterogenouspatternsarelimitedtominingfromdifferentdatadomains.Forexample,inaclinicalap-fromatmosttwodifferentkindsofdata[17