摘要
在知识粒度的基础上,针对决策表提出了相对粒度和属性相对重要性的概念,证明了知识的相对粒度随着知识粒度的增大而单调增加的变化规律,在此基础上提出了一种基于相对粒度的启发式约简算法,以弥补基于正区域的约简方法处理不一致决策表时存在的不足。通过理论分析和实例验证表明,该算法是有效的,且其时间复杂度相对较低。
A relative attribute significance of decision tables, based on the theory of knowledge granularity,was defined by introducing the concept of relative granularity, and the relative granularity＇ s monotonous increasing property with the increase of knowledge granularity was proved, then a heuristic reduction algorithm based on relative granularity was proposed. The algorithm which eliminates the limitation of reduction algorithms based on positive region in dealing with inconsistent decision table, by analyzing essential theory and application examples, is proved as effective, and it＇s time complexity is relatively low.
出处
《计算机科学》
CSCD
北大核心
2009年第3期205-207,共3页
Computer Science
基金
国家自然科学基金项目(69803014
60173058)
河南省高校新世纪优秀人才支持计划(2006HANCET-19)资助
关键词
决策表
知识粒度
相对粒度
属性约简
Decision tables, Knowledge granularity, Relative granularity, Attribute reduction
