With the proliferation of highdimensional data, feature selection has become an indispensable task in the learning process. Attribute reduction is an important feature selection method. To find effective attribute reduction, combining rough set with the ant colony algorithm, we propose a rough set attribute reduction method using the ant colony optimization algorithm. The method firstly updates pheromones, and sets the upper and lower limits for pheromone values. Then the construction scheme of candidate solution is improved according to the addressing mode. Experimental results show that the method has good superiority.