Interactive Network Exploration to Derive Insights: Filtering, Clustering, Grouping, and Simplification

Abstract

Abstract. The growing importance of network analysis has increased attention on interactive exploration to derive insights and support personal, business, legal, scientific, or national security decisions. Since networks are often complex and cluttered, strategies for effective filtering, clustering, grouping, and simplification are helpful in finding key nodes and links, surprising clusters, important groups, or meaningful patterns. We describe readability metrics and strategies that have been implemented in NodeXL, our free and open source network analysis tool, and show examples from our research. While filtering, clustering, and grouping have been used in many tools, we present several advances on these techniques. We also discuss our recent work on motif simplification, in which common patterns are replaced with compact and meaningful glyphs, thereby improving readability.

The Pattern of Policy Change on Disaster Management in China: A Bibliometric Analysis of Policy Documents, 1949–2016

“…In NodeXL, 5 different clusters (subgroups) are marked with different colors. The connotations of keywords are therefore explored through the cluster analysis Shneiderman and Dunne 2012;Himelboim et al 2013), whereas the shifts of policy priorities are observed through analyzing the changes in keywords of high degree centrality.…”

Expressive Authoring of Node-Link Diagrams With Graphies

“…NodeXL [44,45], as a Microsoft Excel plug-in for the creation of node-link diagrams, is accessible to a broader audience. NodeXL lets users specify mappings from data attributes to several visual variables.…”

On Mathematical Optimization for the visualization of frequencies and adjacencies as rectangular maps

“…Fourth, we would like to customize the technique of representing a set of individuals with attached frequencies and proximities as a rectangular map to detect communities in graphs, [21], by analyzing the adjacencies represented in the rectangular map. Finally, our method can also be applied to visualize hierarchical data, in which inside every rectangle a new rectangular map has to be represented by taking into account adjacencies with neighboring rectangles and its inner rectangular maps, [12,27,48]. However, the mathematical optimization treatment of these extensions is not trivial and thus further research is still needed.…”