多源多粒度标记决策系统的最优粒度选择

Optimal Granularity Selection in Multi-source Multi-granularLabeled Decision Systems

多粒度标记决策系统是一种重要的关系数据库,近年来学者们对多粒度标记决策系统的最优粒度选择、属性约简及规则约简展开了研究。然而,目前未见有关多源多粒度标记决策系统的研究。首先,介绍了多源多粒度标记决策系统的概念并给出了多源情况下基本信息粒的表示、上下近似算子和信任函数;然后,讨论了多源多粒度标记决策系统中局部信息源的协调性及其在协调与不协调情况下最优粒度选择的策略,并进一步研究了协调与不协调系统的最优粒度选择策略的关系,给出了同时适用于协调与不协调两种信息系统选择最优粒度的方法。最后,结合局部信息源所选的最优粒度,从不同角度出发给出了乐观、悲观及广义三种选择全局最优粒度的方法。

Multi-granular labeled decision system is an important kind of relational database,the optimal granularity selection,attribute reduction and the rule induction in the multi-granular labeled decision information system are studied by scholars in recent years.However,there is no related report on multi-source situation.First of all,the definition of the multi-source multi-granular labeled decision systems was introduced,and the representation of basic granule,upper and lower approximation operators and belief function under the new system were investigated.Then,the consistence of the each local source among the multi-source multi-granular labeled decision information systems and the strategy of optimal granularity selection in each source were investigated.The relationship of optimal granularity selection strategies between in the consistent and inconsistent systems was further studied,and the optimal granularity selection method for both consistent and inconsistent information sources was presented.Finally,based on the optimal granularity of each local information source,three methods of global optimal granularity selection in the multi-source multi-granular labeled decision information systems were proposed,which are called optimistic,pessimistic and generalized methods respectively.The three methods are designed according to different needs or requirements.