一种新的用于连续值属性离散化的约简算法

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。

A new reduction algorithm for discretization of continuous features

Therearetwoproblemsinthetraditionaldiscretization algorithm when heuristic reduction algorithms are used to find the reduction. One is that the reduction discretizing all attributes may not be found. The other is that the heuristic reduction algorithm needs a great deal of time to get the reduction in the data sets without core. To solve the two problems , a new concept called candidate core is given, which is built on the rough set theory and attribute frequency function, and a new heuristic reduction algorithm based on candidate core (named BCC) is presented. This new heuristic reduction algorithm of BCC can find the reduction of the data sets, which discretizes all attributes. The results of experiments show that the new algorithm can improve the performance of discretization for large data sets.