基于对象分布的连续属性离散化方法  

Method on Discretization of Continuous Attributes Based on Objects Distribution

作　　者：李国和[1] 

机构地区：[1]中国石油大学计算机科学与技术系,北京102249

出　　处：《计算机应用研究》2006年第9期258-260,共3页Application Research of Computers

基　　金：国家自然科学基金资助项目(60473125);中国石油天然气集团公司(CNPC);石油科技中青年创新基金资助项目(05E7013)

摘　　要：为了利用信息系统对象在数据空间中分布,通过对对象的模糊聚类,计算每一类在坐标轴上的统计值。利用正态概率分布特性,引入重叠度和空隙度选取合适的概率,确定最优的区间端点,实现连续属性离散化。In order to use the distribution of objects in information system, statistical values are computed on an axis by fuzzy clustering. By means of the normal distribution, overlap measure and gap measure are introduced, and then a probability is selected suitably, As a result, the terminals of data ranges are obtained optically to realize the discretization of continuous attributes successfully.

关 键 词：信息系统 模糊聚类 概率统计 连续属性 离散化 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]