Attribution reduction is one of the key topics in the field of rough set theory. Based on such theory, the concept of ensemble attribute reduction has been proposed. The ensemble reduction is to divide the sample into multiple decision systems in terms of the decision categories and then calculate them separately. Although ensemble attribute reduction balances the requirements of various decision classes, the corresponding time of attribute reduction is increased. To solve this problem, an attribute reduction acceleration method based on sequential three-way decisions is proposed. The specific steps are as follows: (1) The importance of the attribute in the decision system is calculated. (2) The attributes are divided into three groups in terms of the significance degree of corresponding attribute. Then, the attributes with maximal significance degree are classified into the positive domain, the attributes with zero significance degree are classified into the negative domain, and other attributes will be classified into the boundary domain. (3) The significance degree of the attributes in the boundary domain is  calculated cyclically and the obtained result is divided, until the constraint is satisfied. 8 UCI data sets are selected to conduct experiments in the traditional attribute reduction and ensemble reduction environments, respectively. The experimental results show that, under the premise of ensuring the classification performance, the proposed method can effectively reduce the time of attribute reduction in such two environments.