基于超图的关联规则可视化方法

发布时间：2018-12-21 14:43

【摘要】：数据可视化是数据分析的重要研究方向,在交通、医疗、教育、商务网站、社区、通信等领域广泛应用。数据可视分析旨在以可视交互界面为基础,将人的感知和认知能力融入到数据处理过程中,人脑智能与机器智能优势互补,以获得对数据的洞察力。关联规则挖掘能够从数据中提取有用的、新颖的、属性间存在的有趣的频繁模式或相关关系。对关联规则结果进行可视化,能够使关联分析的结果直观易懂。然而,传统的关联规则可视化方法存在缺乏多模式关联规则展现形式、无法展示关联规则的内部信息以及分布情况等问题,需进一步研究关联规则可视化方法。论文基于超图、超边的相关理论知识,研究基于超图的关联规则可视化方法,设计并实现用户可参与的关联规则挖掘及可视化集成原型系统。首先,阅读国内外可视化相关文献,系统分析超图可视化、频繁项集可视化、关联规则可视化研究现状;其次,详细阐述数据可视分析技术,同时分别概述数据挖掘、可视化、人机交互基本概念以及相关技术;然后,基于超图的相关理论与知识,提出基于无向超边的频繁项集可视化算法和基于有向超边的关联规则可视化算法;最后,基于提出的可视化算法,设计关联规则挖掘及可视化集成原型系统,同时应用于某省全员人口数据集。实验结果表明,所提出的可视化方法具有良好的展示效果。本文的主要工作:1)阐述数据可视分析技术整理数据可视化国内外相关文献,阐述数据可视分析理论知识与技术。首先,根据超图、频繁项集、关联规则可视化文献,总结超图可视化、频繁项集可视化、关联规则可视化研究现状;然后,详细介绍数据可视分析3个主要构成部分:数据挖掘、人机交互、可视化相关技术,为后续研究提供理论基础。2)提出基于超图的频繁项集、关联规则可视化算法针对传统的关联规则可视化方法缺乏多模式关联规则展现形式、无法展示关联规则的内部信息以及分布情况、属性值重要程度突出不明显等问题,提出了基于超图的频繁项集、关联规则可视化算法。首先,基于超图、无向超边的定义及可视化方法,设计基于超图的频繁项集表示模型;结合“圆锥体”布局结构,提出基于超图的频繁项集可视化算法;然后,基于有向超边的定义,给出BF规则图的定义,根据关联规则模式的不同,设计一对一、一对多、多对一、多对多四种模式的关联规则可视化模型;结合“沙漏形”布局结构,提出基于超图的关联规则可视化算法。3)实现基于超图的关联规则挖掘及可视化集成原型系统基于2)提出的可视化算法,结合3D技术和人机交互技术,实现基于超图的关联规则挖掘及可视化集成原型系统。并将系统应用于某省全员人口数据集,展示挖掘出的频繁项集与关联规则。实验结果表明,与传统的频繁项集、关联规则可视化方法相比,论文所提出的可视化方法具有良好的展示效果。
[Abstract]:Data visualization is an important research field in data analysis. It is widely used in the fields of transportation, medical treatment, education, business website, community, communication and so on. The purpose of data visualization analysis is to integrate human perception and cognition into the process of data processing based on visual interaction interface. Human brain intelligence and machine intelligence complement each other in order to gain insight into data. Association rule mining can extract useful, novel, and interesting frequent patterns or relationships between attributes from data. Visualizing the result of association rules can make the result of association analysis intuitionistic and easy to understand. However, there are some problems in the traditional visualization methods of association rules, such as the lack of multi-pattern association rules representation, the lack of internal information and distribution of association rules, and the need to further study the visualization methods of association rules. Based on the theoretical knowledge of hypergraph and hyper-edge, this paper studies the visualization method of association rules based on hypergraph, and designs and implements the association rules mining and visualization integrated prototype system in which users can participate. First of all, read the domestic and international visualization literature, systematically analyze the hypergraph visualization, frequent itemsets visualization, association rules visualization research status quo; Secondly, the technology of data visual analysis is described in detail. At the same time, the basic concepts of data mining, visualization, human-computer interaction and related technologies are summarized respectively. Then, based on the theory and knowledge of hypergraph, we propose an undirected hyperedge-based frequent itemset visualization algorithm and a directed hyperedge-based association rule visualization algorithm. Finally, based on the proposed visualization algorithm, a prototype system of association rule mining and visualization integration is designed and applied to the whole population data set of a province at the same time. Experimental results show that the proposed visualization method has a good display effect. The main work of this paper is as follows: 1) expatiate the data visual analysis technology to sort out the domestic and foreign related documents of data visualization, and expatiate the theory and technology of data visual analysis. First, according to hypergraph, frequent itemsets, association rules visualization literature, summarized hypergraph visualization, frequent itemset visualization, association rule visualization research status quo; Then, three main components of data visual analysis are introduced in detail: data mining, human-computer interaction, visualization related technologies, which provide a theoretical basis for further research. 2) A frequent itemset based on hypergraph is proposed. Aiming at the problems of traditional association rules visualization algorithm, such as the lack of multi-pattern association rules presentation, the lack of the internal information and distribution of association rules, the importance of attribute value is not obvious, and so on. A hypergraph based frequent itemset and association rule visualization algorithm is proposed. Firstly, based on the definition and visualization method of hypergraph and undirected hyperedge, a representation model of frequent itemsets based on hypergraph is designed, and an algorithm for visualization of frequent itemsets based on hypergraph is proposed. Then, based on the definition of directed hyperedge, the definition of BF rule graph is given. According to the different patterns of association rules, the visualization model of association rules is designed for one-to-one, one-to-many, many-to-one, many-to-many, many-to-many patterns. Combined with the "hourglass" layout structure, a hypergraph-based association rule visualization algorithm is proposed. 3) the hypergraph-based association rules mining and visualization integrated prototype system are implemented based on 2). Combined with 3D technology and human-computer interaction technology, an integrated prototype system of association rules mining and visualization based on hypergraph is implemented. The system is applied to the whole population data set of a province to show the frequent itemsets and association rules. The experimental results show that compared with the traditional visualization methods of frequent itemsets and association rules, the visualization method proposed in this paper has a good performance.
【学位授予单位】：河北师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13

【相似文献】