保持不可分辨关系的离散化方法

Discretization method based on indiscernibility

依据不可分辨关系是等价关系的原则将对象分类,对Naive Scaler方法进行补充,对决策表相邻两个实例,在属性值和决策值都不相同时,选取两个属性值的平均值作为断点,连续扫描候选断点,可以使离散化前后的决策表具有相同的不可分辨关系,即原决策表与离散后的决策表是等价的.

Equivalence can be used to classify objects, indiscernibility is one of equivalence, the Naive Scaler is ameliorated. For the two data in the decision table which have different attributes and decision data, the average data is selected as cut-point. Continuous scanning makes the decision table before and after discretization have the same indiscernibility. In other words, the discretization decision table has same indiscernibility as the former decision table, which can bring consistency of theirs.