一种改进的快速数据离散化算法

Improved Fast Discretization Algorithm

提出一种新的基于粗糙集理论的快速数据离散化算法FRSBD(Fast Rough Set based Discretization Algorithm).文章定义了属性决策关系矩阵等概念,证明了一组基于属性决策关系矩阵的断点判定规则的有效性,并基于该新的断点判定规则,实现了决策表中连续属性值的快速离散化.理论分析说明了FRSBD的正确性和有效性,仿真结果表明该算法优于文献报道的同类算法.

This paper presented a fast Rough Sets based diseretization approach, named FRSBD. The approach defines a new conception： The Feature Decision Matrix. A group of diseretization rules based on the Feature Decision Matrix is proved. Based on the novel split rules, we realize fast discretization of continuous features. Theoretical analysis demonstrates that FRSBD is correct and efficient, simulation results show that the performance of it is better than those algorithms reported in literature.