基于规则信息量的规则提取及规则约简

RULE MINING AND RULE REDUCING BASED ON THE INFORMATION OF RULES

在提出规则信息量的基础上,分析了可能性规则的定义,并利用规则信息量对条件属性类和决策属性类间的概率因果关系进行了描述,重新定义了可能性规则。同时,进一步提出了能充分利用决策表信息系统中统计信息的规则提取算法LRE。本文还基于规则信息量得到了一种新的规则约简算法,这种算法是在对原有的粗集规则约简定义进行合理推广的基础上建立起来的,能很好地抵抗数据中的噪声。

Based on the concept of information of rules, the definitions of possible rules are diseased, and then, a new definition of possible rules is given. This new definition helps us to construct the new algorithm for rule mining which named LRE. At last, a new algorithm of rule reducing is given by the improved definition of concept reduction. The algorithms of rule mining and rule reducing put forward in this paper all have the fitness for the noise in database.