pyh3: Scalable and High Performance Graph Visualization in 3D Hyperbolic Space
As the big social networks such as Facebook, Twitter, Instagram etc. are becoming super popular, direct hits for content generating websites such as New York Times and BuzzFeed is taking a smaller and smaller portion of content views. People are more getting used to see these content in social networks in posts shared by friends. According to a news in Business Insider 2013, only 23% of BuzzFeed views are from BuzzFeed.com directly. To analyze how the content is viewed on Internet, we need to trace how a post is shared across network and investigate how the viral content spread. With the spread process into a spanning tree layout, an analytic 3D visualization can effectively deliver such results. The challenge of getting a minimum spanning tree layout from a large dataset comprising the relational information is computationally expensive. The tree layout contains analytic information or results of how information flows through the tree in network, and the layout shape indicates the pattern of information flow. When visualizing a larger tree, a trade-off between overall structure and node details is getting harder to manage. A fish-eye effect visualization is preferred to give a clear overall spanning tree structure and at the meantime, clearly present node details that are currently in focus. H3 (3 dimensional visualization in hyperbolic space) is a good candidate developed by Standford Imagery Laboratory providing a scalable, high performance and interactive graph visualization in 3D hyperbolic space. We found this algorithm is good at tracking and presenting massive amounts of posts sharing across different social networks over time. pyh3 is capable of laying out trees on the order of millions of nodes in a matter of minutes.
|Your Review / Feedback :|
|Your Company :|