Visualizing Very Large Graphs Using Clustering Neighborhoods

7th International Symposium of Hungarian Researchers on Computational Intelligence; pp. 483-494, 2006

The basis of the presented methods for the visualization and clustering of graphs is a novel similarity and distance metric, and the matrix describing the similarity of the nodes in the graph. This matrix represents the type of connections between the nodes in the graph in a compact form, thus it provides a very good starting point for both the clustering and visualization algorithms. Hence visualization is done with the MDS (Multidimensional Scaling) dimensionality reduction technique obtaining the spectral decomposition of this matrix, while the partitioning is based on the results of this step generating a hierarchical representation. A detailed example is shown to justify the capability of the described algorithms for clustering and visualization of the link structure of Web sites.

2001

In several real-life data mining applications, data resides in very high (1000) dimensional space, where both clustering techniques developed for low dimensional spaces (-means, BIRCH, CLARANS, CURE, DBScan etc) as well as visualization methods such as parallel coordinates or projective visualizations, are rendered ineffective. This paper proposes a relationship based approach to clustering that alleviates both problems, side-stepping the "curse of dimensionality" issue by working in a suitable similarity space instead of the original high-dimensional attribute space. The similarity measure used can be tailored to satisfy business criteria such as obtaining user clusters representing comparable amounts of revenue. The clustering algorithm is used to reorder the data points so that the resulting (rearranged) similarity matrix can be readily visualized in two dimensions, with clusters showing up as bands. While such visualization is not novel, the two key contributions of our method are: (i) it leads to clusters of (approximately) equal importance, and (ii) related clusters show up adjacent to one another, further facilitating the visualization of results. Both properties arise from the efficient and scalable top-down graphpartitioning approach used for clustering in similarity space. The visualization is very helpful for assessing and improving clustering. For example, actionable recommendations for splitting or merging of clusters can be easily derived, and it also guides the user towards the right number of clusters. Results are presented on a real retail industry data-set of several thousand customers and products, as well as on clustering of web document collections.

HAL (Le Centre pour la Communication Scientifique Directe), 2017

Many computing applications imply dealing with network data, for example, social networks, communications and computing networks, epidemiological networks, among others. These applications are usually based on multivariate graphs, i.e., graphs in which items and relationships have multiple attributes. Most of the visualization techniques described in the literature for dealing with multivariate graphs focus either on problems associated with the visualization of topology or on problems associated with the visualization of the items' attributes. The integration of these two components (topology and multiple attributes) in a single visualization turns into a challenge due to the necessity of simultaneously representing the connections and mapping attributes possibly generating overlapping elements. Among usual strategies to overcome this legibility problem we find filtering and aggregation that makes possible a simplified representation with reduced size and density providing a general view. However, this simplification may lead to a reduction of the amount of information being displayed, while in several applications the graph details still need to be represented in order to make possible in-depth data analysis. In face of that, we propose ClusterVis, a visualization technique aiming at exploring nodes attributes pertaining to sub-graphs, which are either obtained from clustering algorithms or some user-defined criteria. The technique allows comparing attributes of nodes while keeping the representation of the relationships among them. The technique was implemented within a visualization framework and evaluated by potential users. CCS Concepts •Human-centered Computing ➝ Visualization application domains ➝ Information Visualization.

IEEE Transactions on Visualization and Computer Graphics, 2000

Community structure is an important characteristic of many realnetworks, which shows high concentrations of edges within special groupsof vertices and low concentrations between these groups. Community related graph analysis, such as discovering relationships amongcommunities, identifying attribute-structure relationships, andselecting a large number of vertices with desired structural features and attributes, are common tasks in knowledge discovery in such networks.The clutter and the lack of interactivity often hinder efforts to apply traditional graph visualization techniques in these tasks. In this paper, we propose PIWI, a novel graph visual analytics approach to these tasks. Instead of using Node-Link Diagrams (NLDs), PIWI provides coordinated,uncluttered visualizations and novel interactions based on graph community structure. The novel features, applicability, and limitations of this new technique have been discussed in detail. A set of case studies and preliminary user studies have been conducted with real graphs containing thousands of vertices, which provide supportive evidence about the usefulness of PIWI in community related tasks.

IEEE Transactions on Visualization and Computer Graphics, 2000

One of the most challenging issues in mining information from the World Wide Web is the design of systems that present the data to the end user by clustering them into meaningful semantic categories. We show that the analysis of the results of a clustering engine can significantly take advantage of enhanced graph drawing and visualization techniques. We propose a graph-based user interface for Web clustering engines that makes it possible for the user to explore and visualize the different semantic categories and their relationships at the desired level of detail.

Information Visualization, 2016

Graph visualizations increase the perception of entity relationships in a network. However, as graph size and density increases, readability rapidly diminishes. In this article, we present an end-to-end, tile-based visual analytic approach called graph mapping that utilizes cluster computing to turn large-scale graph (node–link) data into interactive visualizations in modern web browsers. Our approach is designed for end-user analysis of community structure and relationships at macro- and micro scales. We also present the results of several experiments using alternate methods for qualitatively improving comprehensibility of hierarchical community detection visualizations by proposing constraints to state-of-the-art modularity maximization algorithms.

Collection of selected papers of the II International Conference on Information Technology and Nanotechnology, 2016

The article is devoted to the analysis of social networks, which are presented by graphs. The paper presents approaches to the modeling of distributions of social networks as well as the algorithms used for finding communities, as well as accounts that have the greater impact on the community.

SpringerBriefs in Computer Science, 2013

ABSTRACT This work presents a data visualization technique that combines graph-based topology representation and dimensionality reduction methods to visualize the intrinsic data structure in a low-dimensional vector space. The application of graphs in clustering and visualization has several advantages. A graph of important edges (where edges characterize relations and weights represent similarities or distances) provides a compact representation of the entire complex data set. This text describes clustering and visualization methods that are able to utilize information hidden in these graphs, based on the synergistic combination of clustering, graph-theory, neural networks, data visualization, dimensionality reduction, fuzzy methods, and topology learning. The work contains numerous examples to aid in the understanding and implementation of the proposed algorithms, supported by a MATLAB toolbox available at an associated website.

2012

Abstract. This paper describes a graph visualization methodology based on hierarchical maximal modularity clustering, with interactive and significant coarsening and refining possibilities. An application of this method to HIV epidemic analysis in Cuba is outlined. 1