一种新的基于连续属性离散化的属性约简方法  

One method of new attribute reduction based on discretization of continuous attributes

作　　者：胡德敏[1] 冯科峰[1] 

机构地区：[1]上海理工大学计算机与电气学院,上海200093

出　　处：《计算机应用研究》2009年第1期64-65,共2页Application Research of Computers

基　　金：国家自然科学基金资助项目(60573142)

摘　　要：通过将连续属性离散化和属性约简结合起来,首先对连续型的属性列进行离散化,得到新的决策表;然后再对新的决策表作属性约简,解决了属性约简过程中由于不考虑连续属性而无法求出准确约简属性的问题。最后通过具体案例表明了该方法具有较好的实用性、有效性,可以很好地应用在含有大量连续属性的数据挖掘项目中。Through combining the discretization algorithm of continuous attributes with attribute reduction algorithm, this paper firstly did a discretization on continuous attributes columns, and got a new decision-making table. Then did attribute reduction on it, dealt with the problem of getting inaccurate reduction attributes on account of thinking about not continuous attributes during the process of attribute reduction. Finally an illustrated example shows that the way has preferable practicality and availability, can commendably apply to data mining project which contains plentiful continuous attributes.

关 键 词：连续属性 离散化 属性约简 分辨矩阵 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]