数据偏序结构关系中的知识发现可视化方法

摘要

In this paper, the formal concept is first analyzed and partial order structure theory introduced. The deci⁃sion diagram of attribute partial ordered structure ( DDAPOS) , a visualization method of rule extraction and knowl⁃edge discovery based on cognitive principles, is then proposed. After the decision problem is transformed into a de⁃cision pattern information table, the attributes of a research object can be presented in the visualized diagram. This paper introduces the principles, generation algorithm, and application examples of DDAPOS. Experimental results show that the knowledge and rules contained in the data can be represented graphically, and the decision-making rules in the data can be found effectively through analysis of the graph branches, nodes and clusters.%在形式概念分析与偏序结构理论基础上，针对决策模式信息表，提出一种基于认知原理的规则提取与知识发现的可视化新方法———属性偏序决策图。该方法在将决策问题转化为决策模式信息表的基础上，通过研究对象的属性特征，将其表现在可视化图形上，介绍了属性偏序结构图的原理、生成算法及应用实例。实验表明，属性偏序结构图可以将数据中蕴含的知识和规则得以形象地表示，通过对属性偏序决策图支路、节点、簇集的分析可以有效地发现数据中蕴含的决策规则。