推理建模中基于KDD和粗糙集的案例修改

Case Revision Method Based on KDD and Rough Set for Case-Based Modeling

在应用基于案例推理技术进行智能建模时,案例修改后的案例质量好坏直接影响所建模型的精度,但是由于案例修改对领域知识的依赖性很强,采用一般手工案例修改方法无法保证案例修改的质量,即无法保证智能推理模型的精度。基于以上原因,该文提出了一种新的案例修改方法,利用KDD技术,通过有效的多值关联规则挖掘算法从运行数据库中挖掘出案例各属性间的依赖关系,得到案例修改的基本关联规则集,在此基础上利用粗糙集理论对基本关联规则集进行简约,然后根据简约后的关联规则进行案例修改。在线对比实验证明,应用本文方法进行案例修改,提高了修改后的案例质量,从而提高了整体智能推理模型的精度。

The quality of revised ease imposes a direct effect on the model accuracy when the ease - based reasoning（CBR） technology is adopted to make the reasoning model intelligent. Manual ease revision method is being used commonly, and it＇s difficult to guarantee the quality of revised ease for its dependence on domain knowledge, namely it is unable to guarantee the model accuracy. For the reasons mentioned above, a new ease revision method is developed in this paper in which the technology of knowledge discovery in database with effective mining algorithm of quantitative association rules is introduced to find the dependence relation of each attribute from the operational data records and to get the basic association rules set, and then the rough set is used to acquire the reduction rules for revising the eases. An on - line comparison experiment with satisfactory results show that revised ease has high quality by adopting the proposed ease revision model, and the accuracy of CBR can be improved accordingly.