The　performance　of　dimensionality　reduction　on　multiple,　relevant,　and　uncertain　data　is　not　satisfied　by　a　single　method.　Therefore,　in　this　paper,　we　proposed　a　two-stage　reduction　method　based　on　rough　set　and　factor　analysis　(RSFA).　This　method　integrates　the　advantages　of　feature　selection　on　treating　relevant　and　the　advantages　of　rough　set　(RS)　reduction　on　maintaining　classification　power.　At　first,　a　RS　reduction　is　used　to　remove　superfluous　and　interferential　attributes.　Next,　a　factor　analysis　is　utilized　to　extract　common　factors　to　replace　multi-dimension　attributes.　Finally,　the　RSFA　is　verified　by　using　traditional　Chinese　medical　clinical　data　to　predict　patients＇　syndrome.　The　result　shows　that　less　attributes　and　more　accuracy　can　be　expected　with　RSFA,　which　is　an　appropriate　reduction　method　for　such　problems.　©　2012　Springer-Verlag.