摘要
一般启发式值约简算法中启发信息选取不够合理,获取规则的置信度不高,同时还需要多次遍历决策表,时间开销较大。针对上述问题,提出了一种基于加权平均的属性值重要度的概念,并利用受限区分矩阵构造了一种新的值约简方法。该方法无须多次遍历决策表,且不必考虑属性值恢复的问题,大大方便了规则摄取,有效地降低了计算的时间复杂度,且属性值约简更加合理,保证了最后获取的规则具有较高的置信度。最后通过真实的医学数据实验结果表明,该方法具有较好的约简效果。
The common value reduction spends much time on searching the decision tables,and due to heuristic information with no reason in the heuristic algorithm of value reduction,the rule＇s degree of confidence is low.To this issue,this paper presented a concept of attribute value and a new method of value reduction based on discernibility matrix.In this method decision tables is not searched frequently and attribute value＇s resuming doesn＇t need to be considered.It makes incepting the rules more convenient and reduces the time of computing effectively.At the same time,the method makes attribute value reduction with more reason that makes rules with higher degree of confidence.At last,the real physical data experiment proves that the method has a good reduction effect.
出处
《计算机应用研究》
CSCD
北大核心
2010年第11期4098-4100,共3页
Application Research of Computers
基金
国家火炬计划资助项目(2004EB33006)
江苏省高校自然科学指导性计划资助项目(05JKD520050)
关键词
粗糙集
区分矩阵
值约简
决策表
rough set
discernibility matrix
value reduction
decision tables
