In　this　paper,　the　concept　of　the　frequently　covered　points　(FCP)　and　the　infrequently　covered　points　(ICP)　is　presented.　By　means　of　the　cutpoints　sieve　method,　we　can　rapidly　pick　out　the　corresponding　cutpoints　of　ICP,　namely　preferred　cutpoints,　from　all　pending　cutpoints　of　interval　attributes.　And　then,　only　preferred　cutpoints　are　used　for　computing　information　entropy　of　partition　(IEP).　Finally,　the　interval-valued　decision　tree　can　be　built　by　IEP.　The　experiment　indicates　that,　in　general,　this　method　could,　to　a　great　extent,　reduce　the　computational　complexity　of　creation　of　decision　tree,　thereby,　improving　the　efficiency　of　classification.　©　2008　IEEE.