资源描述：
《Using textmining and sentiment analysis for online forums hotspot detection and forecast》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、DecisionSupportSystems48(2010)354–368ContentslistsavailableatScienceDirectDecisionSupportSystemsjournalhomepage:www.elsevier.com/locate/dssUsingtextminingandsentimentanalysisforonlineforumshotspotdetectionandforecastab,c,⁎NanLi,DeshengDashWuaDepartmentofC
2、omputerScience,UniversityofCalifornia,SantaBarbara,USAbReykjavíkUniversity,IcelandcRiskLab,UniversityofToronto,CanadaarticleinfoabstractArticlehistory:Textsentimentanalysis,alsoreferredtoasemotionalpolaritycomputation,hasbecomeaflourishingfrontierReceived1
3、5July2008inthetextminingcommunity.ThispaperstudiesonlineforumshotspotdetectionandforecastusingReceivedinrevisedform8September2009sentimentanalysisandtextminingapproaches.First,wecreateanalgorithmtoautomaticallyanalyzetheAccepted17September2009emotionalpol
4、arityofatextandtoobtainavalueforeachpieceoftext.Second,thisalgorithmiscombinedAvailableonline24September2009withK-meansclusteringandsupportvectormachine(SVM)todevelopunsupervisedtextminingapproach.Weusetheproposedtextminingapproachtogrouptheforumsintovari
5、ousclusters,withthecenterofeachKeywords:Textminingrepresentingahotspotforumwithinthecurrenttimespan.ThedatasetsusedinourempiricalstudiesareSentimentanalysisacquiredandformattedfromSinasportsforums,whichspansarangeof31differenttopicforumsandClusteranalysis
6、220,053posts.ExperimentalresultsdemonstratethatSVMforecastingachieveshighlyconsistentresultsOnlinesportsforumswithK-meansclustering.Thetop10hotspotforumslistedbySVMforecastingresembles80%ofK-meansDynamicinteractingnetworkanalysisclusteringresults.BothSVMa
7、ndK-meansachievethesameresultsforthetop4hotspotforumsoftheyear.Hotspotdetection©2009ElsevierB.V.Allrightsreserved.MachinelearningSupportvectormachine1.Introductionofdataavailableonline.Anothermostrecenttechniquecalledsentimentanalysis,alsoreferredtoasemot
8、ionalpolaritycomputation,hasalwaysIntheInternetandinformationAge,onlinedatausuallygrowsinbeensimultaneouslyemployedwhenconductingonlinetextmining.anexponentialexplosivefashion.ThemajorityofthesewebdataisinThepurpose
