In　this　paper,　first　we　proposed　a　method　of　attribute　reduction　based　on　rough　set　under　fuzzy　equivalent　relation.　We　computed　the　similarity　of　cases　with　the　fuzzy　equivalent　relation,　reduced　the　attributes　by　the　same　fuzzy　equivalent　partitions　based　on　rough　set,　and　then　gave　a　method　of　computing　the　weights　of　the　attributes.　Comparing　with　the　traditional　method　of　attribute　reduction　based　on　rough　set,　more　information　of　the　primary　data　is　held,　and　more　accuracy　of　the　attribute　reduction　is　enhanced　by　our　method.　©　2007　IEEE.