Unified Mapping of Social Networks into 3D Space

Social network visualization focuses on nodes lacking inherent layout semantics. Many previous mapping algorithms locating graph data into physical space are based on assumptions that space is in two-dimensions and that the visualization of huge graphs can be realized by dividing them into subgraphs independently for easier exploration. However, when physical space extends to three dimensions, the automatic layout becomes a problem. Different from sparse graphs, nodes in social networks may be closely connected to each other, which raises another problem that how to explore them in undivided ways. For these reasons, the primary motivation for this work is to provide a unified framework optimized for social network a 3D exploration by combining techniques widely used in data mining and computational graphics. In the framework, we borrow octree structure to solve nodes self-adapting layout problem in 3D space and propose an algorithm to map a social network into an octree without losing connection information after space division, and define two practical operations (shifting and shading) to satisfy the details-on-demand requirement of data visualization. By building a prototype system on Java 3D, the experiment visualizes the social networks extracted from real organizations and demonstrates the flexibility of the framework. Our methods produce a satisfying performance.