与特征选取和离散化集成的决策规则挖掘方法

Decision Rules Mining Method by Integrating Feature Selection and Discretization

利用粗糙集理论可以从已知数据中挖掘决策规则 .对于连续取值的特征属性必须先对其离散化 .从给定的特征属性集合中去除冗余的特征属性 ,选取有用的属性子集有助于简化决策规则 .我们利用基于信息熵的规则不确定性量度函数构造了一个决策规则挖掘的遗传算法 ,将规则挖掘与特征选取和连续属性的离散化集成在一起 .

Decision rules can be mined from given data using rough set theory. The continuous features must be discretized. Removing the redundant feature attributes and selecting the useful feature subset can simplify the decision rules. We construct a genetic algorithm for decision rules mining integrated by feature selection and discretization using an entropy based uncertainty measure. The usefulness of the proposed method is demonstrated by the experimental results.