针对传统聚类算法受数据空间分布影响大且效率较低的问题,提出一种应用粗糙集理论的聚类算法。以信息表中条件属性与决策属性的一致性原理为基础,以数据超立方体、信息熵实现数据属性约简和离散化。在此基础上,利用集合特征向量加法法则运算,只需扫描一次信息表就可实现对数据对象的聚类划分。实验结果表明该算法是有效可行的。 Aiming at the problem that the traditional clustering algorithm is greatly affected by the spatial distribution of data and the efficiency is low, a clustering algorithm based on rough set theory is proposed. Based on the consistency principle of conditional attributes and decision attributes in the information table, data attribute reduction and discretization are realized by hypercube data and information entropy. On this basis, using the collection of eigenvector addition algorithm, only one scan of the information table can be achieved on the data object clustering. Experimental results show that the algorithm is effective and feasible.