Graph Drawing

The world contains vast amounts of networked data formed from “connections” among “things.” Examples of large-scale networks of data include link relationships between websites, social network friendships, and joint authorship relationships for academic work. One field for visualizing information handling such network data is called a graph drawing. In network data, a “thing” is called a node, and a “connection” is called a link. Visualizing the network data makes it possible to gain an awareness of various aspects such as how individual nodes are connected, where important nodes are within the network, and what type of group (clusters or communities) the nodes constitute. In addition, such awareness supports effective decision-making in marketing, product planning, policy formation, and other factors.

There are various graph drawing methods depending on the type of data and the desired type of visualization and analysis. At Koyamada Laboratory, we develop new graph drawing methods, design and develop visual analysis systems, and carry out applied research using these methods and systems. Four studies related to a graph drawing are introduced below.

1. Visualization of Cognitive Structures using Evaluation Grid Method

The evaluation grid method is a qualitative survey method using semi-structured interviews, and is widely used in product planning among other areas. The evaluation grid method obtains a network structure of people's cognition through interviews. Visualization of the cognitive structure extracted by the evaluation grid method was for a long time carried out by hand. In this research, we extended the Sugiyama framework, which is a graph drawing method, and developed our own graph drawing method suitable for cognitive structure visualization [1]. In addition, based on the above, we developed a cognitive structure visual analysis system that integrates a network analysis method, and effectively analyzed the evaluation structure [2].

2. Visualization of Nematode Embryo Phenotype Characteristic Network

Unraveling the mechanisms behind the occurrence of living organisms is an important issue as a foundation for realizing regenerative medicine for human beings in the future. Through this research, we have engaged in an analysis of the causal relationships between the amounts of phenotypic characteristics measured during the development processes of nematode embryos. These causal relationships can be visualized using the graph drawing method; however, because the connections between the nodes are dense, a large amount of edge crossing occurs, making it difficult to read and understand the connections between nodes. In this research, to generate an easily readable visualization diagram, we developed a new edge concentration algorithm that replaces subgraphs with densely connected edges with graphs of a simpler structure [3]. In addition, applying the edge concentration technique allows us to tackle the identification of phenotypic characteristics and groups of genes causing abnormalities in such characteristics.

3. Development of Structural Equation Modeling Support System

Structural equation modeling (SEM) is a useful statistical analysis method for analyzing the causal relationships among variables. When analyzing causal relationships, there is a need for processes to improve the models in terms of observation variables and latent variables, as well as the addition of paths. Through this research, we will visualize the information necessary for improving SEM models in an integrated manner, and attempt to construct an environment in which analysts can effectively analyze causal relationships.

4. Visualization of Culture in Academic Fields

In recent years, interdisciplinary integrated research through which multiple academic fields collaborate with each other has attracted attention as a means of solving various social problems. To promote interdisciplinary integrated research, it is important that the researchers in each academic field understand each other well. Through this research, based on the results of questionnaire surveys regarding the values and behaviors of researchers, we used a network to visualize how to approach research for each academic field. Based on this research, we aim to promote a mutual understanding and interdisciplinary integration among different academic fields.

Research Achievements


Comprehensive Visualization

Causal Relationship Visualization