In　order　to　overcome　the　defect　of　high　time　complexity　and　wide　search　space　for　finding　the　minimal　attribute　set　during　attribute　reduction.　In　this　paper,　the　reduction　of　attributes　is　considered　as　a　special　optimization　process　by　ant　colony　algorithm　displayed　good　performance　in　solving　complex　problem　of　combinational　optimization.　First,　an　attribute　is　abstracted　as　a　node　and　the　lease　combination　of　these　nodes　is　found　which　can　take　place　in　all　attribute　node　s　but　not　change　the　degree　of　classified　roughness.　Base　on　these　points,　rough　set　theory　and　ant　colony　algorithm　are　combined　according　to　the　problem　of　ant　colony　algorithm　with　little　information　pheromone　and　slow　convergence　rate.　The　attribute　core　is　determined　through　the　correlative　algorithms　of　rough　set　theory,　which　can　be　used　as　initial　node　of　ant　colony　algorithm.　Finally,　the　least　attribute　set　is　scanned　by　means　of　the　search　capacity　of　ant　colony　algorithm.　Theory　analysis　and　experimental　results　show　that　the　algorithm　proposed　in　this　paper　is　feasible　and　effective.