Web访问挖掘研究

《Web访问挖掘研究》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、东南大学博士学位论文Web访问挖掘研究姓名：宋爱波申请学位级别：博士专业：计算机应用技术指导教师：董逸生20030801于Petri网方法。这两种方法，都不需要用户的参与协作，完全是自动的。基于Petri网方法，具有直观性和不需要反复计算推荐集合的优点，而且Petri网能很自然的描述web中的多内容并发显示和并发浏览。关键词数据挖掘web访问挖掘个性化站点管理Petri网IIResearchonWebAccessMiningAbs仃actToday,theWorldWideWebisrapidlyemergingasanimportantmedium

2、forthedissemination，exchange，andgettingofinformation．Accordingtomostpredictions，themajorityofhumaninformationwillbeavailableontheWebintenyears．Thesehugeamountsofdataraiseagrandchallenge，namely，howtoturntheWebintomoreusefulinformationutility．Atpresent，themaintoolsofgettinginform

3、ationarestillsearchengines．Today’Ssearchengines，however，areplaguedbythefollowingfourproblems：thelowprecisionproblem；thelowrecallproblem；Alimitedqueryinterfacethatisonlybasedonkeyword—orientedsearch，andhavenofunctionofcustomizationtoindividualusers．Theseproblems，inturn，Canbeattr

4、ibutedtothefollowingcharacteristicsoftheWeb．Firstandforemost，theWebisahuge，diverseanddynamiccollectionofinterlinkedhypertextdocuments．Furthermore，itiswidelybelievedthat99％oftheinformationontheWebisofnointerestto99％ofthepeople．Second，exceptforhyperlinks，theWebislargelyunstructur

5、ed．Finally，mostinformationontheWebisintheformofHTMLdocumentsforwhichanalysisandextractionofcontentisverydifficult．Therefore，itisnoteasytoovercomeallproblemscausedbysearchengines．Inthisthesis，weusewebaccessminingtodiscoverlaserbrowsingpatternssuchasaims，interests，andpreferences．

6、Thenthesepatternsareutilizedinimprovingthestructureofwebsitesandthemannerofwebservice．Thus，wecallhelpusersgettingwhattheyneedmoreeasilybypersonalizedinformationserviceandautomatedsiteadministration．ItisalsoimportantforE—commercetominethewebaccessdata．Thesignificanceliesinimprov

7、ingthecustomerrelationshipmanagement，assistinginmakingdecisionandsecuritymanagement，andhelpingmerchandisersimplementingaone-to-ODemarketingstrategy．Thedissertationiscomposedofthefollowingparts：(1)Wediscussvariousproblemsmetduringdatapreparinginwebaccessmining，thengiveasimplemet

8、hodtoidentifyuseraccesstransactionsaccordingtohostaddr