A Dynamic Approach for Visualizing Local and Global Information in Geo-spatial Network Visualizations

Abstract

Geo-spatial network visualization is a fusion of visualization techniques and geographic information science, and has progressed rapidly in recent years. However, challenges still remain regarding the clarity of high density node-link graphs.

We present a dynamic approach for revealing the underlying information in locally cluttered areas while maintaining global edge trends around them, by creating and positioning multiple radial layouts within a geo-spatial connected graph. Moreover, two time series data-flow visualization approaches at both local and global scales are proposed respectively, namely: the peapod model and the hub model. The peapod model focuses on data flows within the local area while the hub model addresses the relations between groups of nodes across the graph. The approach is further refined by leveraging integrated circular layouts, stacked graphs, and word cloud techniques in several complementary views to support the exploration of categorical and aggregated data.

In addition to adhering to emergent design standards for geo-spatial visualization, the computational complexity and quantitative performance analysis on three different datasets were subsequently conducted to examine the scalability of the visualization model. The simulation results show that the main algorithms in our approach are able to achieve acceptable performance in real world test cases. Finally, our model's effectiveness are demonstrated by two significant case studies in different application fields, namely: US aero traffic network and GitHub collaboration graph. The case studies show that insightful knowledge can be discovered by the utilization of the dynamic approach and the support of various complementary features.