摘要
提出一种基于邻域系统的决策表近似算法,用于数据挖掘预处理阶段的数据压缩。该方法以代表元素代替若干相近元素,有效地压缩了原始决策表的对象个数,同时保证决策表本身的判断能力基本不变。对本算法与聚类算法的关系进行了讨论,指出文中提出的近似算法所完成的功能不能用聚类算法替代。
This paper presents a neighborhood system based approximation algorithm for decision tables, which may be used to reduce the object number of a decision table in the preprocessing period of data mining. When one object is selected to represent its similar objects, the similar objects are deleted from the decision table. The object number is greatly reduced while the decision ability of the table is not significantly decreased. The comparison between clustering algorithm based on density function and the approximation algorithm shows that this approximation algorithm can't be replaced by cluster algorithm. '
出处
《计算机应用》
CSCD
北大核心
2003年第12期1-2,6,共3页
journal of Computer Applications
基金
国家自然科学基金项目 (6 0 2 750 2 2
6 0 2 0 3 0 1 1 )
关键词
邻域
代表元素
近似
neighborhood
representatives
approximation
