Big graph visual analytics

Today’s graph data problems are increasingly multi-faceted and multi-disciplinary in nature. Many cutting-edge R&D efforts in the graph and visualization community are conducted independently in disparate domains such as bioinformatics, cybersecurity, and predictive machine learning. Although technology transfers in big graph visualization are recognized and growing, many R&D problems and challenges remain unsolved today.

This special issue features six articles previously presented but not published at the Institute of Electrical and Electronics Engineers (IEEE) Workshop on Exploring Graphs at Scale (EGAS) co-located at IEEE VIS 2015 in Chicago, IL, USA. We invited keynote speaker Dr Paul Burkhardt of US National Security Agency, as well as authors of five presentations at EGAS to further revise and extend their presentations for possible inclusion in this special issue. All six submissions have gone through the standard Information Visualization’s double-blind review cycle before their acceptance by the journal.

In “Graphing Trillions of Triangles,” Burkhardt discusses why computing on graphs at increasing scales requires some re-thinking about how we approach these problems by using the problems of counting and listing triangles. And, of course, visualizing the results of graph computations is an important aspect of this large graph landscape. The study and results presented in this article represent the largest and fastest effort ever reported in graph triangle listing.

In “Visual Analytics of Large Dynamic Digraphs,” Burch explores a temporal sequence of a graph over the same vertices and shows how to highlight similarities and differences in the graph as time proceeds. His approach is demonstrated using two case studies: a software call graph changing over time as the application development proceeds and a social network dataset representing face-to-face contacts during a conference recorded by radio-frequency identification (RFID) badges.

An ego-network is the region around one person in a social graph. Liu et al. of “egoComp: A Node-link Based Technique for Visual Comparison of Ego-network” explore techniques for visual analysis of node differences between two ego-networks. The authors present three different graph layout techniques and explore the strengths of each algorithm for the analysis.

Visualizing clusters of vertices in a graph that represent possible communities is an important problem. In “Graph Mapping: Multi-scale Community Visualization of Massive Graph Data,” Crouch et al. present a tile-based visual analytic approach to showing community structure and relationships at varying scales. Experimental results are given showing the impact to community comprehensibility of proposed constraints to state-of-the-art modularity maximization algorithms.

Large graph visualization applied to specific application domains allows for creative solutions to real problems. Leonard and Madduri in “Graph-based Visual Analysis for Large-scale Hydrological Modeling” present a hydrological modeling problem that benefits from graph visualization, and the data required for this analysis are definitely large. The nearly 1-billion edge and more than a half-billion vertex graph they used represent data fused from heterogeneous national data products covering the continental United States.

In “Graph Query Algebra and Visual Proximity Rules for Biological Pathway Exploration,” Wu et al. introduce a pathway graph query algebra and some visual proximity rules on which to develop an exploratory multi-view visualization environment supporting the study of biological paths in large and complex pathway datasets. Inspired by “query-by-example,” their system interprets relationship queries given by several user-supplied examples and reacts to visual proximity rules to provide the biologists with an intuitive and dynamic system for exploring pathway graphs.

These articles present solutions for some of the emerging problems that challenge conventional approaches in computation and interaction, including the latest web-scale graphs with over a trillion edges. Although the data size that seems big today will be different from what will seem big in a few years, we hope that the articles of this special issue will stimulate further research and development that will affect future thinking in the area of big graph visual analytics for years to come.

We would like to thank our authors for their contributions and the journal reviewers for their helpful and informative comments. We also appreciate the generous encouragement from Information Visualization by giving us such a valuable opportunity to publish the special issue.