面向可视化展示的图布局评估方法研究

发布时间：2018-03-28 04:17

本文选题：可视化　切入点：图布局　出处：《华东师范大学》2017年硕士论文

【摘要】：随着信息化时代的到来,社交网络、金融交易、生物网络等复杂网络每天都产生着海量大规模的数据,对复杂网络的数据进行分析和挖掘可以发现出许多有价值的信息。但是复杂网络的节点数目众多,边的关系也非常复杂,传统的表格等文字表现方式无法满足用户对网络结构数据理解、挖掘的需求。网络图可视化以图形的方式把网络图隐藏在数据里的信息快速直观地加以展示,从而使人们更加深入地理解关系数据,已成为近年来网络图数据分析的主要手段。现有的图布局可视化算法很多,各种图布局算法在不同方面有优有劣,图布局的质量在很大程度上影响了用户对图结构的理解。一些可视化手段单纯从技术角度进行革新,而忽略了人类的认知规律和心理映像,导致了很多可视化效果难以被用户理解和接受的现象。从人类的感知出发,对不同图布局的直观性和有效性进行科学的评价能够帮助用户选择出符合心理映像的图布局,使其一眼看穿网络数据背后的关系特征。因此,针对网络图的特点,进行考虑用户感知属性的图布局评估和分析,具有很重要的理论与应用价值。本文提出了两种对图布局进行评估分析的方法。首先针对用户需要关注拓扑重要的节点的情形,结合视觉因素对图布局效果的影响,提出了一种基于视觉感知的图布局评估方法。具体地,基于用户实验分析影响节点的视觉重要性的因子并设计模型对节点的视觉重要性进行量化,通过对比节点的视觉重要性和拓扑结构重要性之间的差异来评估图布局的优劣,一方面借助热力图来直观地展示两者之间的差异,另一方面通过计算精确得到差异值,选择出更符合人类的视觉感知的图布局算法。同时,本文还提出了客观的图布局整体质量评估方法。首先建立图布局主观质量评分的数据库,请用户通过主观实验对不同图布局效果进行评分;接下来对于每一个图布局,计算各个可能影响到图布局效果的指标;最后,我们把图布局中影响布局效果的指标作为自变量,用户对不同图布局的主观评分作为因变量,建立二者的回归模型,通过回归模型对一个图布局进行整体质量评分。本文中的两种图布局评估方法具有不同的适用场景,通过文中提出的图布局评估方法选取出符合用户心理映像的图布局,使其深入地理解图中的特征信息。实验分析显示出我们的方法与实际情况切合度很好。
[Abstract]:With the advent of the information age, social networks, financial transactions, biological networks and other complex networks produce massive and large-scale data every day. A lot of valuable information can be found by analyzing and mining the data of complex networks, but the number of nodes in complex networks is large, and the relationship between edges is very complex. The traditional representation methods such as tables can not meet the needs of users to understand and mine the network structure data. The information hidden in the data can be displayed quickly and intuitively by the visualization of network graph. So that people can understand the relational data more deeply, which has become the main means of network graph data analysis in recent years. There are many existing visualization algorithms for graph layout, and all kinds of graph layout algorithms have some advantages and disadvantages in different aspects. The quality of the layout greatly affects the user's understanding of the graph structure. Some visualization methods are innovated purely from a technical point of view, while ignoring the cognitive laws and psychological images of human beings. Many visualization effects are difficult to be understood and accepted by users. From the perspective of human perception, scientific evaluation of the intuitiveness and effectiveness of different map layout can help users to choose the map layout that conforms to the psychological image. Therefore, according to the characteristics of the network diagram, the graph layout evaluation and analysis considering the user-aware attributes are carried out. This paper presents two methods to evaluate and analyze the layout of the graph. Firstly, considering the situation where the user needs to pay attention to the important nodes of topology, combined with the influence of the visual factors on the effect of the layout of the graph, this paper proposes two methods to evaluate and analyze the layout of the graph. In this paper, a visual perception based graph layout evaluation method is proposed. Specifically, based on user experiments, the factors affecting the visual importance of nodes are analyzed and a model is designed to quantify the visual importance of nodes. By comparing the difference between the visual importance of nodes and the importance of topological structure, we evaluate the layout of the map. On the one hand, we can show the difference between the two by means of thermodynamic diagram, on the other hand, we can get the difference value by calculating accurately. This paper also proposes an objective method to evaluate the overall quality of map layout. Firstly, the database of subjective quality score of map layout is established. Users are asked to score different layout effects through subjective experiments; next, for each diagram layout, we calculate each index that may affect the layout effect. Finally, we take the index that affects the layout effect as an independent variable. The subjective score of different graph layout is taken as dependent variable, the regression model of them is established, and the overall quality of a graph layout is evaluated by regression model. The two evaluation methods of graph layout in this paper have different applicable scenarios. Through the graph layout evaluation method proposed in this paper, the graph layout which accords with the user's psychological image is selected, and the feature information in the graph is deeply understood. The experimental analysis shows that our method is very suitable to the actual situation.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：O157.5

【参考文献】