Abstract: This paper puts forward an efficient algorithm for reduction of attribute in decision systems. A relatively reasonable formula measuring attribute significance is discussed and a quick method to compute partition is proposed. Then a quick algorithm for reduction of attribute is obtained. Compared with those existed algorithms, its flexibility has been increased because calculating the important value of unimportant attributes repeatedly can be avoided by removing unimportant attributes gradually from the search space. The theoretical analysis shows that this algorithm is much less time complexity than those existed algorithms. A real example and experimental results demonstrate its feasibility and effectiveness, respectively.