基于遗传算法的数量属性离散化算法

A GA-based algorithm for discretizing values of quantitative attributes

提出了在没有任何领域知识可供借鉴的情况下,基于聚类思想,利用遗传算法对数量型属性进行离散化的新算法———遗传C均值算法.该算法利用遗传算法具有全局寻优的特性,对训练样本根据其每一属性值进行聚类,将样本划分为不同的类,从而为每一属性找到其值的最佳分割点.然后,对不同类赋以不同的编码.该算法的优点是能得到最优的离散化结果.在VC++6.0环境下实现了该算法.仿真实验证明该方法有效解决了利用粗糙集理论进行分类规则挖掘时,数量型属性的离散化问题.

A cluster-based algorithm for discretizing values of quantitative attributes is presented, called genetic-C means algorithm, which works without any experienced knowledge about special field. The algorithm takes the advantage of genetic algorithms for global optimization, and can get the best cuts for any one attribute via clustering examples. And it is implemented in VC(++)6.0. Computer simulations prove its validity for finding the best cuts for the values of quantitative attributes.