Abstract: Objects in actual decision tables are often changed dynamically. Aiming at this problem, the concept of a simplified decision table is introduced firstly, and a large number of repeated objects are deleted. A simplified matrix based on the positive region is constructed, which can reduce the searching space. What's more, it is theoretically analyzed that the definition of attribute reduction based on the simplified matrix is equal to that based on the un-simplified matrix. The proposed method scans only once the simplified matrix and it can be effectively utilized to the old attribute reduction. On this condition, an efficient algorithm for dynamic attribute reduction based on a matrix was designed. Finally, an example and its experimental comparison were employed to illustrate the efficiency and feasibility of the proposed algorithm.