資源描述:
《鄰域粗糙集及其基于鄰域粗糙集分類算法》由會(huì)員上傳分享,免費(fèi)在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、摘要經(jīng)典粗糙集理論是定義在等價(jià)關(guān)系的基礎(chǔ)上,只能處理名義型數(shù)據(jù),對(duì)數(shù)值型數(shù)據(jù)必須離散化后才可以處理.由于實(shí)際應(yīng)用中的數(shù)據(jù)往往是數(shù)值的,而且測(cè)量時(shí)本身帶有誤差,這給直接應(yīng)用帶來(lái)不便.連續(xù)數(shù)據(jù)經(jīng)過(guò)離散化處理可能會(huì)丟失重要信息并且不同的離散化策略會(huì)影響最終的處理效果.因此,數(shù)值型信息系統(tǒng)的屬性約簡(jiǎn)是近年來(lái)粗糙集理論研究的熱點(diǎn)之一.本文通過(guò)引入鄰域關(guān)系粗糙集模型來(lái)進(jìn)行數(shù)值屬性約簡(jiǎn)和分類器構(gòu)造.首先研究了經(jīng)典粗糙集理論及其性質(zhì),在此基礎(chǔ)上,給出了鄰域關(guān)系粗糙集理論及其相關(guān)性質(zhì):其次,利用鄰域粗糙集中的相關(guān)性質(zhì)和鄰域關(guān)系矩陣
2、的性質(zhì)對(duì)文[32]中屬性約簡(jiǎn)算法做了改進(jìn),并提出了一種基于決策表的新的快速屬性約簡(jiǎn)算法:然后,針對(duì)傳統(tǒng)KNN分類方法在分類時(shí)只考慮最近鄰的樣本點(diǎn)信息,而未考慮該樣本點(diǎn)的近鄰信息,提出了一種考慮近鄰信息的基于鄰域粗糙集的分類算法:進(jìn)一步的考慮到樣本的某個(gè)屬性的取值特別大時(shí),分類時(shí)會(huì)將其他取值小的屬性的信息淹沒(méi),于是提出利用壓縮映射將數(shù)值較大的某些屬性壓縮到一個(gè)合理范圍,給出了基于壓縮映射的鄰域粗糙集分類方法;最后,對(duì)各種分類方法進(jìn)行了比較研究,并通過(guò)實(shí)例驗(yàn)證本文分類算法能夠快速高效地進(jìn)行分類.關(guān)鍵詞:鄰域粗糙集:屬
3、性約簡(jiǎn):鄰域關(guān)系矩陣:KNN分類AbstractTheClassicalRoughSetTheory,definedonthebasisoftheequivalencerelation,Callonlydealwithnominaldata.NumericaldatamustbediscreditedbeforetheyCallbehandled.Asthedatainthepracticalapplicationusuallynumerical,andnotSOaccurateitself嘶msomeerror
4、sintheprocessionofthemeasurement,whichhavecausedinconveniencetothedirectapplication.Inthiscondition,Discriminationofnumericdatawillleadtothelossofsomeimportantinformation,anddifferenttreatmentsandstrategieswillalsoaffectthefinalresults.Therefore,theNumericAtt
5、ributeReduetionofinformationsystemsisoneofthehottestissuesonRoughSetTheoryinrecentyears.ThispaperintroducestheNeighborhoodRoughSetmodeltothenumericalAttributeReductionandClassifica=tionStructures.Firstly,ithasstudiedtheclassicalRoughSetTheoryanditsproperties,
6、andgiventheNeighborhoodRoughSetTheoryanditsrelatedproperties.Secondly,theuseofRoughSetsrelatedtoneighborhoodcharacterandthenatureoftheneighborhoodrelationmatrixisofferedtoimprovetheAttributeReductionAlgorithminthepaper[32],andanewandfastAttributeReductionAlgo
7、rithmbasedondecisiontableisproposed.ThenbyanalyzingthetraditionalKNNClassificationMethodforclassifyingthesamplewhichonlyconsidersthenearestpointofinformation,withouttheaccountofthesamplepointsoftheneighborinformation,thekindofneighborhood-rough-set-basedclass
8、ificationalgorithmisbroughtabouttoembracetheneighborinformation.FurtherthisarticlehasconcludedtheNeighborhoodRoughSetClassificationMethodbasedonCompressionMappingtocompressthecertainprope