一种基于类别属性关联程度最大化离散算法

针对现有离散化算法难以兼顾计算速度和求解质量这一难题,提出一种新的基于类别属性关联程度最大化监督离散化算法.该算法考虑了类别、属性值的空间分布特征,根据类别与属性之间的内在联系构造离散化框架,使离散化后类别和属性的关联程度最大.实验结果表明,基于类别属性关联程度最大化离散算法在保证计算速度的前提下能有效提高分类精度,减少分类规则个数.

A class-attribute interdependency maximization based algorithm for supervised discretization

Considering that the existing discretization algorithms do not give simultaneously attention to evolution speed
and solution’s quality, a new class-attribute interdependency maximization based algorithm for supervised discretization
is proposed in this paper. The algorithm considers the distribution of both class and continuous attributes, and according
to the underlying correlation structure of them, the discretization scheme is constructed which can maintain the highest
interdependence between the target class and all the discretized attributes. The experiment results show that, with a reasonable
execution time, the proposed algorithm can improve the accuracy of the classification result and reduce the number of
classification rules.