【摘要】 在分布式数据库基础上,对各个站点的数据库使用粗集的方法进行挖掘,产生各个站点的规则,将这些规则库组合起来,产生一个全局的规则库,从而可以为管理者提供决策的依据。但是,产生的全局库中往往会有不一致的规则出现,一方面,是由于数据库本身的不一致等原因,导致了规则的不一致;而另一方面,则由于各个站点都追求规则的简洁性,使得直接从全局数据提取的规则不矛盾,在分布式环境下却得到矛盾的规则。对于第一种情况,可以通过现有的增加规则的可信度等方法加以避免;而对于第二种情况,本文提出了三种算法来解决了这个问题,并对这三种算法的效率、得到的一致规则的长度等问题进行详细的分析,说明了算法3是一种高效实用的算法。另外,本文对局部站点的核和全局站点的核的关系进行了研究,并证明了一个定理。更多还原

【Abstract】 On the basis of distributed database,data mining with rough set theory is employed in each site. A global knowledge base comes into being when rules generated from each site are united, which provide a basis for decision-making. However there are always two reasons for some inconsistent rules in the global knowledge base: on the one hand from the data of database itself which are inconsistent and lead to inconsistent rules. On the other hand from the contradiction under distributed environment, because of the pursuit of rules brevity in each site, which may not occur between the rules extracted directly from the whole data composed of data in each site. For the first one, the inconsistency can be avoided by increasing the confidence of current rules. For the second one, this paper proposes three algorithms to solve the problem. The efficiency of three algorithms and the length of the consistent rules obtained are discussed in detail that shows the algorithm three is the best one. Additionally, a theorem is proved for finding the relation between the local core and the global core of rough set theory.更多还原