Casual Exploration of Large Time-Dependent Bipartite Graphs

Bipartite graphs are typically visualized using linked lists or matrices. However, these classic visualization techniques neither scale well with the number of nodes nor is it possible to convey temporal developments of the graph. Interactive visualization techniques for large bipartite graphs typically introduce special encodings that reveal very limited information on the first glance.

Dynamic Bipartite Cluster Flows (BiCFlows) is a new casual exploration interface for large, time-dependent bipartite graphs for a general audience. We use two different clustering techniques to build a hierarchical aggregation of nodes and edges supporting different exploration strategies. Aggregated nodes and edges are then visualized with classic linked lists and nested time series visualizations. We demonstrate the usefulness of the technique using two data sets: a public database of media advertising expenses of public authorities in Austria and author-keyword co-occurrences from the IEEE Visualization Publication collection.

How it works

The bar chart on the top left contains different types of legal bases or conferences, respectively. Selecting one or multiple items in this bar chart filters the visualization. Below, there is a time line. Adjusting the time range below the time line also filters the visualization. To drill further into the individual clusters, click on the purple or green bars, respectively. The thin bars on the sides can be used to navigate back to the overview. Hovering a label highlights all its connections. Labels can also be selected from the two text lists on the bottom left.


You can watch a tour through the two data sets first using BiCFlows, then using Dynamic BiCFlows (i.e., also investigating the temporal component):
video [mp4, 35 MB]


BiCFlows has been presented at the International Symposium on Big Data Visual and Immersive Analytics in October 2018 in Konstanz, Germany:
BDVA paper [PDF, 1.8 MB]

Source Code

Created by Daniel Steinböck at TU Wien, extended by Wolfgang Knecht and Johannes Sorger at the Complexity Science Hub Vienna:
available at GitHub


This work was conducted by Daniel Steinböck and supervised by Manuela Waldner and Eduard Gröller from the Institute of Visual Computing and Human-Centered Technology at TU Wien and later extended by Wolfgang Knecht and Johannes Sorger from the Complexity Science Hub Vienna.

This work was funded by the Austrian Science Fund (FWF): T752-N30.