semi supervis approach rapid reliabl label larg data set gy rgy simon vipin kumar zhi li zhang paper propos method label data set carri semi supervis manner user guarante qualiti label scheme assum class heurist identifi instanc class heurist assum reason perform cover instanc class perfectli reliabl assum infal expert manual label instanc aim algorithm exploit cluster structur problem predict imperfect heurist limit perfect label provid expert classifi label instanc data set guarante precis specif user class precis attain algorithm allow classifi instanc dontknow algorithm evalu number instanc label expert number dontknow instanc global coverag achiev qualiti label kdd cup network intrus data set instanc manag label instanc guarante nomin precis confid expert label instanc expert label instanc manag guarante nomin precis label data provid case studi appli scheme label network traffic collect larg campu network 