Abstract: In this paper, granularity was described by utilizing the equivalence relation in Rough sets, whose expression in information system and decision table was given by combining rough set with granularity calculation method. And using the attribute importance value in information decision table as heuristic information, Tabu search was carried out on the combination of relative reduction numbers. This method can avoid useless attributes selection, effectively wipe off attributes that can be omitted and reduce the search space, so as to ensure the efficiency of the algorithm.

Key words: knowledge granularity, importance value, rough set, text expression, heuristic information