针对在 Nguyen和 Skowron的离散化算法中进行启发式约简时会出现某些属性不能进行离散化问题 ,以及在无核数据集中启发式约简算法计算量比较大等问题 ,在粗糙集理论和属性频率函数的基础上给出一个新概念——候选核 ,并提出一种新的用于连续值属性离散化的约简算法——基于候选核的启发式约简算法 (简称 BCC)。该算法可以寻找到能对所有属性进行离散化的约简。实验表明 ,所提出的 BCC算法能提高大数据集的离散化效果。 In the case of heuristic reduction in Nguyen and Skowron’s discretization algorithm, some properties can not be discretized, and the computational complexity of heuristic reduction algorithm in non-nuclear data sets is relatively large. In the rough set theory and Based on the frequency function, a new concept, candidate kernel, is proposed. A new reduction algorithm for discretization of continuous value attributes is proposed, which is called Heuristic Reduction Algorithm (BCC) based on candidate kernel. The algorithm can find reductions that can discretize all attributes. Experiments show that the proposed BCC algorithm can improve the discretization effect of big data set.