基于粗糙集理論與支持向量機(jī)的數(shù)據(jù)挖掘方法算法研究.pdf

1、Y1213777論文題目(中文)—盞王擔(dān)撞墓壁監(jiān)當(dāng)童掛由堇扭i蝗———_婪監(jiān)控?fù)?dān)友溘簋迭匠蠱—一論文題目(外文)——衛(wèi)mL叢血凹出盤血四￡壁盟g血一——旦墼魚!＆曼鯉曲上韭d地籃曲趟一————』ⅢmmL!盟地￡叢塹M盟￡——一研究生姓名———————二己盂—韭—————一學(xué)科、專業(yè)———————數(shù)芏L直監(jiān)璀睹L———一研究方向———J蜘塹越生堡監(jiān)丑煎握控地——一學(xué)位級(jí)別———————』互_————————一導(dǎo)師姓名、職稱—————』耋赴

2、盥蛙睦L————一論文工作起止年月—』迦立生』。且至上四衛(wèi)生上旦—一論文提交日期—墊業(yè)業(yè)衛(wèi)jL論文答辯日期———型地L生—衛(wèi)—且——一學(xué)位授予日期摩蠆淪蟲～晰。萄究英文摘要ABSTRACTThisthesisstudiessomekeytechnologyquestionsindataminingbasedontongllsettheoryfirstlyItiswellknownthatthereareusuallymuchredund

3、antdatainlargeknowledgerepositoryThesedatawastethestoragespaceanddisturbmakingdecisionInthethesis。theknowledgereductionisstudiedfromequivalentrelationofatrributesystemangleofview，fromthedependentlevelofattributeandimport

4、anceofattributeangleofviewfromthediseemibilitymatrixangleofviewfromtheviewpointofinformationtheoryByexperimentalstudies，fivewaysreducedtheaRdbuteswereobtainedMeanwhilewediscoverthatthechangingtendencyoftheinformmionentro

5、pyisnonrigorousmonotonicallydecreasingincomentropy’whenthenumberoftheattributesisincreasingInfactthesortedboundaryisveryaccuratewhenitisclassifiedonthebasisofclassicalroughsetstheoryAlthoughaccuracyimprovesgreatlyinthose

6、waysfortheclassificationrecognitionitsfaulttoleranceandtheserviceabilityofmodelisverypoorForremovingthedefects，thisthesishasstudiedvariableprecisionroughselstheoryanditsreductionThenitstudiedmodelingprincipleofclassifica

7、tionmethodofSVMandregressionanalysisofSVMandstudiedtheirsphereofapplicationandproblemsolvingAtonetime，itw笛discoveredthatitsmeritsishiddendangerindataminingIftherealenoiseorcontradictoryinformationresultspredictionbasedon

8、smallsamplesetwillbegreatlyinfluencedBeforeforecastingandclassifyingofSVM，thesequestionshavebeenfoundandforeclosed，whichispreciselywhatroughsetstheoryhasadvantageUponthaLbasedonadvantageandvirtueoftheroughsetstheoryandSV

9、Mmethods，thispaperstudiedhowtocombinetwomethodsandobtainedproceduremethodswhichorganicallycombineRSl丫RoughSetsTheory)withmorethanoneclassificationlearningmachineofSVMAlso。ithasgivenamethodthatthroughthemconstructmorethan