属性分类的多层次形式概念模型及动态算法研究

Research on Multi Level Formal Conceptual Model and Dynamic Algorithm Based on Attribute Classification

在传统形式概念中,属性是以单粒度单层次进行分析,然而现实中树形结构的属性分类是普遍存在的.针对属性分类的不同粒度层次关系,本文提出了属性分类的多层次形式概念模型,分析形式概念在不同层次泛化空间下相关性质.结合粒计算的思想,本文首先提出了属性分类的属性泛化与细化方法,属性泛化约简形式概念,属性细化提高分析精度.然后由属性分类层次的变化,给出了一种动态的形式概念构造算法,该算法通过自学习的方式对原有知识加以使用,它不仅继承了目前渐进式算法的优点,还实现形式概念自身数据的变化.最后通过实例和仿真实验说明本文所提方法的有效性.

Most traditional studies on formal concept analysis, the attribute is single granularity and single hierarchy. However, attribute taxonomic are usually predefined in applications and represented by hierarchy trees. Attribute classification for different granularity lev- els, this paper proposes a multilevel formal concept analysis based on attribute classification, analysis of the related properties of the formal concept in different levels. Based on granular computing,First,a method of attribute generalization and refinement based on for- mal concept analysis is proposed, attribute generalization reduction concept and attribute refinement to improve accuracy. Second,based on the change of attribute classification level,this paper presents a dynamic formal concept construction algorithm, the algorithm uses the original knowledge by self-learning, It not only inherits the advantages of the incremental algorithm, but also can deal with the change of the data of the formal concept. Finally, example and experimental results demonstrate that the proposed method is effective.