Many knowledg¢ discovery tools have beell developed using data millillg , む he integratillg te〔thnologies of machille learging , daむ abase , statisti 〔 : s and others .、 Ve have been collst, 1 ' uctill9 ' 匚 mondo ゴ ' seal ℃ h systems based on extei ' ided assoeiation rules .111 this paper , " ・ c) discusse the experimelltal results of text minillg applied for web llyper texts, 1NSPEC da. t, abase , をmd lllagazh1 {} s a 皿 d articles index data iu the National Diet l・ ibvary. First of all , we

more »

... ress abou も the ef且cient 帥 rategies ill order to derive association Tules . Next) we discuss the relation beし ween the threshold values al訌d asso (: iat , ion rule 呂 , and we f { ⊃cus on し 11e techniqties of ROC ( Reciever Operating Characteristics )grap } I to evaluate tlle characteristics of derived rules 、 By usillg the ROC coIlvex ftlll rnethod , we can estimate aPpropria し e 寸hl ・ eSh 。ld valites 匕〔コ derive aSSOC 五 atiO 【 1 rUleS fOr keywOrdS.