Visualization Of Association Rules Based On Metagraph

Visual knowledge discovery of association rules is a nontrival process of identifying valid, novel, potentially useful, and ultimately understandable patterns using visualization techniques. As the brain of human being has a strong ability of pattern recognition, combining visual knowledge discovery process and visualization technology would be usefull to fulfill the advantage of the human visual cognition, as well as reducing the burden of understanding, giving new insights from the abstract data, and contributing to the knowledge conversion and creation. Process model has an important guiding significance for the implementation of the project. Meanwhile, information visualization and knowledge visualization are very important parts in visual knowledge discovery process of association rules. However, current researches focusing on the visual knowledge discovery process model of association rules are rare, the existing visualization methods of association rules also have many problems. In order to achieve a more valuable result, there is a need to combine knowledge discovery process model with great visualization methods. So this dissertation did the following work:1) Proposing a more practical, more comprehensive application model of visual knowledge discovery process, which is called V-KDDM. By introducing the theory of data-information-knowledge-wisdom hierarchy and integrating the thinking pattern when human brain managing knowledge, it is easier for decision-makers to grasp a macro of the whole knowledge discovery and data mining implementation process; and by combining with a variety of visualization techniques, using the advantage of human’s visual system, the model gradually implements transformation from data to information, from information to knowledge, and from knowledge to wisdom; and by viewing the entire V-KDDM model as a process of human interaction and giving fully play to the initiative of the user, ultimately achieve knowledge sharing.2) The traditional association rule information visualization approaches are mainly orienting to expert users while ignoring the normal user’s perception and comprehension ability. Meanwhile when the number of rules increasing, edges and nodes prone to overlap, as well as result in reducing the performance and readability of rule representation. This dissertation proposes a novel form of information visualization method based on S-C metagraph to show one to one, one to many, many-to-many association rules. This dissertation gives the basic definition of S-C metagraph and the model showing association rules using S-C metagraph, also gives the properties and derivation process of S-C metagraph for visualizing association rules. Furthermore, the concept relationship is added to the S-C metagraph based association rule information visualization method and constructe Vis-Meta graph, which is used for association rule knowledge visualization. This dissertation gives the relevant definitions of Vis-Meta graph and Vis-Meta graph knowledge visualization method of association rules, difines the conceptual relationship in Vis-Meta graph for knowledge visualization, and gives the knowledge visualization algorithms based on those, which converting tacit knowledge to explicit knowledge.3) Designing and achieving the prototype system of visual knowledge discovery of association rules. By applying demographic data of a province to the prototype system, the effectiveness of the association rule information visualization method and knowledge visualization method are verified, and the values of V-KDDM model are also proved.