一种改进的连续属性模糊离散化方法  

One Improved Continuous Attribute Fuzzy Discretization Method

作　　者：焦冬艳[1] 于津[2] 张华[3] 

机构地区：[1]广东理工职业学院计算机系 [2]汕头大学工学院 [3]河南财经政法大学计算机学院

出　　处：《现代计算机》2011年第1期20-22,28,共4页Modern Computer

摘　　要：连续属性离散化是知识系统中的一个重要环节,一个好的离散化方法能够简化知识的描述和便于对知识系统的处理。而求取连续属性值的最优断点集合是一个NP难题。提出一种连续属性模糊离散化的Norm-FD方法:根据正态分布特点采用正态离散化算法(Norm-D算法),使其离散结果达到需要离散区间数,根据属性值和与其相邻的区间关系将具体属性值用F-Inter算法转化为用隶属度、分区号和偏向系数三个参数表示。Discretization of continuous attribute values is an important part of the knowledge system.A good discretization method can simplify description of knowledge and facilitate the treatment to knowledge systems.Obtaining the optimal breakpoint set is a NP problem.Proposes a fuzzy discretization of continuous attributes-Norm-FD method,according to the characteristics of normal distribution,normal dispersion method as Norm-D algorithm is used to achieve the required results of the discrete intervals.Then the attribute values are converted to the three parameters for the membership,the partition number and the bias factor by using F-Inter algorithm according to attribute value and its adjacent interval between the attribute values,which completed a fuzzy discretization of continuous attributes.

关 键 词：模糊集 连续属性离散化 正态分布 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]