基于決策樹算法的文本圖像分類方法研究.pdf

1、分類號(hào)：密級(jí)：公開學(xué)號(hào)：2009230單位代碼：10407碩士學(xué)位論文論文題目:基于決策樹算法的文本圖像分類方法研究研究方向:數(shù)據(jù)挖掘?qū)I(yè)名稱:計(jì)算機(jī)應(yīng)用技術(shù)研究生姓名:黃細(xì)妹導(dǎo)師姓名、職稱:喻金平教授2012年6月4日江西贛州IIAbstractAlongwiththerapiddevelopmentofmoderncomputerwktechnologyimageaccesstechnologytheInterimagedatain

2、creasedsharply.Thereisaphenomenonthattheimagedataisquiterichbuttheusefulimageinfmationispo.Tosolvetheproblematechnologyisneededtominethepotentialusefulinfmationintheimage.Thusthetechniqueofimagemininghasemerged.Imagemini

3、ngtechnologyisdesignedtoextractmuchpotentialinfmationinimageplayaroleofimageinfmationdecisionmaking.Nowadaysthoughimagemininghasbecomeahotspotitisstillayoungtechnologyneedstobefurtherstudied.Textimageisakindofspecialimag

4、ethatcontainstextinfmation.Itplaysanimptantpartineverydaylifehasbecometheresearchobjectofdatamining.Toclearlyunderstthetextinfmationintextimageitisnecessarytoclassifythetextimage.Inthispapertheimageminingclassificational

5、githmwasintroducedatfirst.Thenthestudiesondecisiontreeclassificationalgithmtheimagefeaturewereconducted.Finallyitfocusesonthetextimageclassification.Themainresearchcontentsresultsareasfollows:(1)AnimprovedID3algithmbased

6、onanewattributesioncriterionwasproposed.ThetraditionalID3algithmusestheinfmationgainasattributesioncriterionbutthiscriterionhastwodeficienciesoneisthelongalgithmexecutiontimetheotheristheattributeioncriteriontendstovalue

7、ofmultipleattributes.TosolvethetwodeficienciesasimplifiedmethodbasedonequivalentinfinitesimaltheattributeimptancemethodthatchangedbytheimptancedegreeofattributeswereusedtoimprovetheID3algithm.Finallyitmakesanewattributes

8、ioncriterion.(2)Anindepthstudyonthethreekindsoftextimagefeatureswasconducted.Thetextimagefeatureswasobtainedfromimagehistogramgraylevelcooccurrencematrix.Thefeaturevectsusedtodistinguishtextimageswereobtainedbycomparingt

9、extimageacteristicsincludingmeanvarianceskewnesscontrasthomogeneityenergycrelation.(3)Atextimageclassificationmethodbasedonthecombinationofunderlyingimagefeaturewasproposedinthispaper.Themethodusesthetextimagefeaturevect

10、thatobtainfmthetextimagefeatureextractionprocesstoclassifytextimageswithID3algithmC4.5algithmimprovedID3algithm.Theclassificationmethodwasverifiedbyexperimentstheresultshowsthatitownsahighperfmance.InthispaperimprovedID3