Using　a　multi-objective　linear　programming,　we　develop　an　approach　to　set　the　decision　power　for　the　multiple　fuzzy　rules,　properly　in　a　highly　fuzziness　event　and　earnings　manipulation,　whose　degree　of　membership　is　difficult　to　observe.　We　use　the　proposed　method　to　detect　the　uncovered　earnings　manipulators　during　the　period　2001–2010　from　the　companies　listed　at　Shanghai　Stock　Exchange　and　Shenzhen　Stock　Exchange.　The　recognition　rate　for　in-sample　test　is　78.4　%,　and　the　corresponding　rate　for　out-of-sample　test　is　76.9　%.　©　Springer-Verlag　Berlin　Heidelberg　2014.