一种基于粗糙集启发式的特征选择算法

本文基于粗糙集中关于非精确集和精确集理论思想,提出了一个新的特征度量指标,即相对互信息比RMI,由此,设计了一种基于粗糙集的启发式特征选择算法MRMI-UC。首先利用可辨识矩阵,计算出条件属性相对于决策属性的核,以核形成当前候选特征子集作为基准点,以最大化相对互信息和不确定性系数为原则,筛选剩余特征。通过对比实验,结果表明,本文提出的算法在多数情况下能够得到较优的特征子集,算法是有效的,切实可行的。

A Novel Feature Selection Heuristic Algorithm Based on Rough Set Theory

In this paper, a new feature measurement RMI (Ratio of Mutual Information)is presented based on the concept of rough set theory about certain set and uncertain set. Then a novel heuristic algorithm, MRMI-UC (Algorithm based on Maximal Ratio of RMI and Uncertainty Coefficient), is proposed for Feature Selection based on rough set theory. Firstly, the Core is obtained by discernible matrix and formed as a candidate feature subset. With the starting point of Core, the rest features are filtered iteratively to maximize both RMI and Uncertainty Coefficient. Finally the algorithm is tested on the UCI datasets, experiment results show that MRMI-UC is feasible and can find a good feature subset in most cases.