摘要
通过引入属性值通配符,进而将规则表示为带通配符的样本——规则样本,并对粗集的知识约简方法稍作修改,使得加入新样本后更新规则库时能够充分利用已经获得的规则,尽量减小待约简的决策表的数据量,避免每次从庞大的原始决策表开始约简,从而加快更新速度,减小计算量,只作少量修改,而不必再从头约简.最后结合一个实例阐明了该方法的基本思路.
By defining the attribute-value wildcard, each rule is expressed in a form called 'Rule Sample', which contains wildcards (as compared with the original 'experimental sample',with a definite value for each attribute ). Then with the classical RS rule-extracting algorithm slightly modified, rules can be easily refreshed whenever new samples are added to the decision table, because the application of wildcard can help to avoid processing from the original decision table(usually contains extremely large amount of da ta) every time. Therefore time and efforts are saved greatly when the rules need to be refreshed. An example is given to demonstrate the main idea of this improved algorithm.
出处
《武汉大学学报(理学版)》
CAS
CSCD
北大核心
2002年第3期370-374,共5页
Journal of Wuhan University:Natural Science Edition
基金
国家自然科学基金资助项目(69896240)
关键词
粗集
增量式规则提取算法
决策表
rough sets
decision table
rule
incremental learning