面向?qū)＠麛?shù)據(jù)清洗和分析的自動分類方法研究.pdf

1、X1406603中圖分類號魚三』3：l；!￡!!!UDC學校代碼苧Q9壘!密級壘五中國科學技術(shù)信息研究所碩士學位論文StudyonPatentCategorizationforDataCleanandAnalysis作者姓名塹壁塞一導師姓名韭坦捏塞監(jiān)學位類別籃理堂學科專業(yè)一一值握堂學號809010612職稱副盟宜旦盟宜旦學位級別亟研究方向讓簋扭籃J墾堇壟中國科學技術(shù)信息研究所論文提交日期(2008年05月)StudyonPatentCa

2、tegorizationforDataCleanandAnalysisAbstractInpacewiththerapiddevelopingofS&T’allofthecountries，industriesandcompaniesintheworldhavestrongrequirementofcompetitivetechnologyintelligenceandtechnology’StendencyMeanwhile，when

3、wecarryoutsomeprogramssuchaLstechnologymonitoringandtechnologyforecastingwefinditverynecessaryforUStomakeadeepanalysisaboutpatentTherefore，itisourfocalpointtodesignsuitablealgorithmforpatentanalysisBYmeansofdeepstudyabou

4、tthecurrentsituationofpatentanalysis，wefindthattherearetwopuzzlesOneisthatwhenwetrytocleanpatent’SdataitalwaystakesUSalongtimeTheotheriswhenwewanttoanalyzesomepatentdata，wefineithardforUStoclassi句thernInordertosolvethese

5、problems，thisPaperbringsanideaforewordthatwecandeviseacategorizationalgorithmtolookupsomeisolatedpointinthedatasetwhichhelpsUStocleandataandtoclassifythemForthisreason，thisthesisgetsadeepstudyabouttextandpatentcategoriza

6、tionThen，borrowingideasfromCNKI’Sword’Ssimilarityalgorithmthisstudyputsforewordallautomaticpatent’ScategorizationalgorithmthatisbaseduponatreelikeIPCstructureAfterthatthisstudyimplementsthealgorithmbyusingClanguageFurthe

7、rmore，wediSCUSSsomewaystoselectproperparameterinthealgorithmFinallybyusingsometrainingdatathataregottenfromaprogramoftheMinistryofScienceandTechnologywetrytoevaluatethisalgorithmandtheresultisthatwhenitisusedindatacleani

8、tsprecise’rateis079080itsrecall’Srateis091092andwhenitisusedindataanalysisitsprecise’rateis099itsrecall，Srateis098ThatismeansthealgorithmwhichdesignedbythispaperCanuseintechnologymonitoringandtechnologyforecastingandmake