摘要
在知识发现流程中,分类规则是主要的挖掘任务之一。针对传统的基于统计分析的挖掘算法在保证知识的有趣性方面的缺陷,提出了利用演化计算这种智能计算模型的全局搜索特性和完全适应值导向特性来进行分类知识的自动挖掘和处理,不需要先验知识,以确保知识的有趣性。提出了用IF-THEN这种高层次的知识表示形式来提高知识的可理解性。并给出了个体表示,遗传操作和适应值评估等几个在演化算法中起重要作用的成分的设计原则和方法。
Mining the classification rules is the main task in the course of Knowledge discovery in Database(KDD).In this paper,we focus on mining classification rules based on evolutionary algorithm(EA)to guarantee that the mined knowledge is interesting for users because EA has many advantages like fitness-driven and global searching.It is em-phasized that using high-level knowledge representations like IF-THEN(prediction)rules is helpful for knowledge com-prehensibility.The principle of how to analyze and design the key components in GA(individual representation,genetic operation,fitness measurement )is presented.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第2期13-15,31,共4页
Computer Engineering and Applications
基金
国家自然科学基金(编号:69635030
60073043
70071042)
关键词
数据挖掘
分类规则
演化算法
有趣性
适应值
Data mining,Classification rule,Evolutionary algorithm,Fitness measurement
