The　task　of　Public　Opinion　Detection　and　Tracking　monitors　the　hot　topics　in　the　forums,　BBS　and　so　on.　It　collects　the　latest　news　and　views　and　then　the　classifying　and　clustering　algorithms　are　applied.　Finally,　the　monitoring　results　are　presented　to　the　end　users.　This　task　has　given　the　significant　impact　on　the　large　numbers　of　internet　users.　The　effective　features　in　the　story　document　are　selected　and　we　adopt　the　vector　center　model　to　represent　the　text　document.　The　clustering　algorithm　merges　the　story　to　the　corresponding　cluster.　We　adopt　three　types　of　performance　evaluation　metric　and　that　is　FValue,　the　entropy　value　and　the　Square　Error　Function　JC.　We　analyze　the　distribution　of　the　monitoring　results　on　different　forums.　The　results　indicate　that　our　approach　is　effective.　The　future　research　will　focus　on　the　detection　of　the　timing　news　reports,　and　extracting　the　unique　features　to　study　the　adaptive　detection　model　and　strategies.　©　2012　ACADEMY　PUBLISHER.